Cement and Concrete Research 34 (2004) 1275–1296

Review article

The confused world of sulfate attack on concrete

Adam Neville*

A and M Neville Engineering, 24 Gun Wharf, 130 Wapping High Street, London, E1W 2NH, UK

Received 23 March 2004; accepted 8 April 2004

Abstract

External sulfate attack is not completely understood. Part I identifies the issues involved, pointing out disagreements, and distinguishes

between the mere occurrence of chemical reactions of sulfates with hydrated cement paste and the damage or deterioration of concrete; only

the latter are taken to represent sulfate attack. Furthermore, sulfate attack is defined as deleterious action involving sulfate ions; if the reaction

is physical, then, it is physical sulfate attack that takes place. The discussion of the two forms of sulfate attack leads to a recommendation for

distinct nomenclature. Sulfate attack on concrete structures in service is not widespread, and the amount of laboratory-based research seems

to be disproportionately large. The mechanisms of attack by different sulfates—sodium, calcium, and magnesium—are discussed, including

the issue of topochemical and through-solution reactions. The specific aspects of the action of magnesium sulfate are discussed, and the

differences between laboratory conditions and field exposure are pointed out.

Part II discusses the progress of sulfate attack and its manifestations. This is followed by a discussion of making sulfate-resisting concrete.

One of the measures is to use Type V cement, and this topic is extensively discussed. Likewise, the influence of w/c on sulfate resistance is

considered. The two parameters are not independent of one another. Moreover, the cation in the sulfate salt has a strong bearing on the

efficiency of the Type V cement. Recent interpretations of the Bureau of Reclamation tests, both long term and accelerated, are evaluated, and

it appears that they need reworking.

Part III reviews the standards and guides for the classification of the severity of exposure of structures to sulfates and points out the lack of

calibration of the various classes of exposure. A particular problem is the classification of soils because much depends on the extraction ratio

of sulfate in the soil: there is a need for a standardized approach. Taking soil samples is discussed, with particular reference to interpreting

highly variable contents of sulfates. The consequences of disturbed drainage of the soil adjacent to foundations and of excessive irrigation,

coupled with the use of fertilizer, are described. Whether concrete has undergone sulfate attack can be established by determining the change

in the compressive strength since the time of placing the concrete. The rejection of this method and the reliance on determining the tensile

strength of concrete because of ‘‘layered damage’’ are erroneous. Scanning electron microscopy (SEM) should not be the primary, and

certainly not the first, method of determining whether sulfate attack has occurred. Mathematical modeling will be of help in the future but, at

present, cannot provide guidance on the sulfate resistance of concrete in structures.

Part IV presents conclusions and an overview of the situation, with consideration of future improvements. Appendix A contains the

classification of exposure to sulfate given by various codes and guides.

D 2004 Elsevier Ltd. All rights reserved.

Keywords: C Sulfate; Attack; Soil; Groundwater; Type V

This paper is in four parts. Part I establishes what is

meant by sulfate attack. Part II discusses the progress and

consequences of sulfate attack. Part III reviews the testing

of aggressive soils and waters, and also testing for the

purpose of establishing whether sulfate attack of concrete

in the field has occurred. And Part IV presents an

overview.

0008-8846/$ – see front matter D 2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.cemconres.2004.04.004

* Fax: +44-207-265-1087.

1. Part I: What is meant by sulfate attack?

1.1. Introduction

I have toyed with using the word ‘‘topsy-turvy’’ instead

of ‘‘confused’’ because topsy-turvy is defined in the Shorter

Oxford Dictionary as ‘‘utterly confused’’. This is how the

situation with respect to sulfate attack seems to me, but

perhaps, topsy-turvy is not a proper adjective for a research

journal. It is not that our knowledge of the various phe-

nomena and the actions involved have never been clarified;

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961276

they are understood, at least at the engineering level, but the

disputes in the last 10 years or so have resulted in many

assertions. These, each taken separately, are not necessarily

wrong, but taken together, result in a thoroughly confused

situation.

I suppose that at the outset, I should justify my attempt to

find some general strands in the issue of sulfate attack of

actual structures: I am an engineer and not a scientist, and

my view is that the behavior of structures in the field is the

professional concern of an engineer.

I am not alone in emphasizing the need for engineering

judgment and, therefore, for an engineer to be prominently

involved in studies of sulfate attack. DePuy [1] of the

Bureau of Reclamation wrote that, in certain circumstances,

‘‘it becomes increasingly important that the engineers have a

good understanding of the chemical reactions, factors gov-

erning the rate of reaction, and mechanisms causing deteri-

oration in concrete’’.

Nor am I alone in pointing out that laboratory studies to

date are inadequate to translate into an understanding of

field behavior. DePuy [1] wrote: ‘‘Most of the knowledge of

sulfate attack has been developed from laboratory studies

involving relatively simple chemical systems of pure mate-

rials, yet in actual practice the situation is more

complicated. . .’’. And, like me, DePuy [1] feels that ‘‘al-

though sulfate attack has been extensively investigated, it is

still not completely understood’’.

1.2. Scope of the paper

I do not claim that the present paper will make the

situation crystal clear. I am unable to achieve this, in part,

because I am a civil engineer, and understanding sulfate

attack involves not only civil engineering and construc-

tion practices, but also chemistry, soil engineering, and

field and laboratory testing. It is not easy to get people

from all these disciplines to work together. Worse still,

there is sometimes a tendency for any one specialist

to opine on phenomena outside his or her specialism.

Unfortunately, some expert witnesses are tempted to ex-

press opinions outside their own field of expertise, fearing

that admitting ignorance might detract from their stand-

ing, or else, they are inveigled into opining in alien

fields.

Of course, I, too, must try not to fall into that trap, and

yet, I am writing on the broad issue of external sulfate

attack. So let me reassure readers that I intend to limit

myself to identifying the issues involved and to pointing out

the disagreements. However, of course, this will not prevent

me from expressing my views, where I have some confi-

dence in doing so.

‘‘Chemists Should be Studying Chemical Attack on

Concrete’’ is the title of an article by Hime [2], published

in Concrete International in April 2003. And yet I, an

engineer, am writing on a topic that concerns chemical

action on concrete. Why?

First of all, Hime [2] himself says that chemists have not

been ‘‘doing the chemistry’’ of chemical attack.

Second, I am not expressing any views on chemical

reactions, although I shall need to refer to some of them.

Third, the real significance of chemical reactions be-

tween sulfates and concrete is in the damage to concrete.

Mere chemical changes that do not affect the health,

performance, or durability of concrete are of very limited

interest to those concerned with real structures in the field,

as distinct from laboratory specimens under artificial con-

ditions. I hasten to add that laboratory research is important

and essential to advance our knowledge, but the practical

significance of that knowledge needs to be examined in

engineering terms.

This brings me to the fourth issue: How widespread is

sulfate attack? And also, how significant are the con-

sequences of the reactions of hydrated cement paste with

sulfates?

The present paper is not concerned with internal sulfate

attack or with thaumasite attack, which is a separate form of

sulfate attack [3]. Nor does the paper specifically deal with

sulfate attack on concrete in consequence of oxidation of

iron sulfide, such as pyrite, in the soil, which is largely an

acid attack because sulfuric acid is formed. This can have

very serious consequences, but allowing concrete to come

into contact with soil containing pyrite is simply bad

practice and can be avoided or prevented.

1.3. Definition of ‘‘attack’’

At this juncture, it may be appropriate to define the

meaning of the term attack. Usually, I find the New Shorter

Oxford Dictionary, published in 1993, to be a good starting

point. The Dictionary says, ‘‘a destructive action by a

physical agency, corrosion, eating away, dissolution’’. I

see the adjective ‘‘destructive’’ as being crucial; in other

words, an action by itself does not necessarily represent

attack. From an engineer’s point of view, what matters is

what has happened to the concrete: an action that does not

result in deterioration or in a loss of durability is not an

attack.

My view is supported by some researchers in the field of

thaumasite attack. Bensted [4] says, ‘‘It is important that

thaumasite formation and thaumasite sulfate attack are

properly differentiated’’. This distinction is of vital impor-

tance. The mere fact of a given chemical reaction occurring

is not evidence of damage and of consequential loss to the

owner of the given concrete. The distinction of Bensted [4]

is equally applicable to sulfate attack that does not result in

the formation of thaumasite.

Further support to my views is given by Harrison [5],

who says, ‘‘When the analysis of concrete reveals a high

sulphate content this does not necessarily indicate any

deterioration although conversely, loss of strength or visible

deterioration accompanied by a high sulphate content would

be evidence of sulphate attack’’.

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1277

In addition, Skalny et al. [6] emphasize that ‘‘the pres-

ence of ettringite per se is not a sign of sulfate attack’’. On

the other hand, they refer to efflorescence as ‘‘virtually

observable damage’’ and describe efflorescence as being

‘‘predominantly sodium sulfate; occasionally also sodium

chloride, magnesium sulfate, other salts’’ [6]. They say that

efflorescence represents severe chemical attack, the word

chemical being italicized in their book.

My view that efflorescence per se is not a proof of

damage has been expressed earlier [7]. Mehta [8] also says

that efflorescence ‘‘should not cause any damage’’, except

under certain circumstances.

1.4. Definition of sulfate attack

I have already expressed the view that attack on concrete

occurs only if concrete experiences deterioration or damage.

This still leaves open the question of what kind of mecha-

nism needs to be involved for the action to be described as

sulfate attack. There are two schools of thought.

One school considers sulfate attack to have taken place if

sulfates are involved, regardless of the mechanisms acting.

The other school limits the concept of sulfate attack to the

consequences of chemical reactions between sulfate ions

and hydrated cement paste, so that chemical changes in the

paste take place. However, if sulfates interact with cement

and cause damage to it, but the action is physical, and a

similar action can occur with salts other than sulfates, then

the damage is considered to be physical attack or physical

sulfate attack.

I am aware of the fact that chemical changes are

accompanied by physical changes, and I hesitate to enter

into the niceties of science. Why then am I raising the issue?

Indeed, it could be argued that nomenclature is of little

significance. However, in my opinion, in a large and

complex field, it is helpful if everybody uses the same

terminology when referring to a given phenomenon.

Furthermore, I believe that there is a difference between

chemical and physical attack in that the chemical attack

involves necessarily the sulfate ion. On the other hand,

physical attack involves crystallization of salts, of which

one example is sulfate, but it is a topic of wider interest with

respect to concrete and, also, rocks. In consequence, in this

paper, the term sulfate attack will be limited to chemical

attack, but I shall also consider briefly the physical sulfate

attack.

I am far from being alone in holding the abovementioned

views. For example, the Concrete Society Technical Report

on Diagnosis of Deterioration of Concrete Structures [9]

recognizes ‘‘physical salt weathering’’ as a phenomenon

separate from sulfate attack, regardless of whether the salts

are sulfates or not.

I would also like to quote Collepardi [10] who, in a state-

of-the-art review of delayed ettringite attack on concrete,

considers sulfate both of internal and external origin but

explicitly excludes ‘‘damage which is specifically attributed

to the physical sulfate attack such as crystallization of water-

soluble sulfate salts’’. Mehta [8], in a paper subtitled

‘‘Separating Myths from Reality’’, says: ‘‘Surface scaling

of concrete due to the physical salt-weathering attack is

being confused with the chemical sulfate attack. Salt weath-

ering is purely a physical phenomenon that occurs under

certain environmental conditions when any porous solid,

such as brick, stone, or poor-quality concrete, is exposed to

alkali salt solutions including sulfates (but not necessarily

limited to sulfates)’’. I am not sure that the qualification

‘‘poor quality’’ applied to concrete is appropriate; rather, the

openness of the surface texture is relevant, and this could

exist with some types of finishes of any concrete. Mehta [8]

reports that ‘‘the physical attack associated with salt crys-

tallization was not limited to alkali–sulfate solutions; for

example, it occurs with alkali carbonate solution. . .’’.I have to record that opposite views are also held. For

example, Skalny et al. [11] say that ‘‘to draw a distinction

between ‘physical’ and ‘chemical’ processes does not serve

a useful purpose and will only serve to further confuse

engineers’’. Being an engineer, I find this solicitude touch-

ing, but I do not believe that we, engineers, are already

confused and have to be saved from being further confused!

I would like to emphasize that the issue of chemical or

physical attack is not just semantic because the consequen-

ces of physical attack manifest themselves in a different way

from chemical attack. Moreover, prevention of the two types

of attack is likely to be different.

1.5. Physical sulfate attack

A prevalent form of physical attack is the reversible

change of anhydrous sodium sulfate (thenardite) into deca-

hydrate (mirabilite). If crystallization takes place in the

pores at or near the surface of concrete, large pressure

may develop, with consequent deleterious action. Hime

and Mather [12] say that ‘‘precipitation from supersaturated

solution may be required’’. The term ‘‘may’’ shows that the

phenomena involved are not yet fully understood.

The problem of physical salt attack has been studied for a

very long time. In 1929, Lafuma [13], in France, reported

his experiments on the transformation of sodium sulfate

decahydrate into anhydrous form of that salt and pointed out

the significance of the temperature of 33 jC, above which

the anhydrous salt is formed in a solution with a low

apparent volume. While the minutiae of the papers of

Lafuma [13] may, or may not, be correct—and they are

not within my field of expertise—I have studied them

because they recognize clearly that the actions involved

are physical and that the chemical action of sulfate ions does

not enter the picture.

In 1939, the crystallization of salts in porous materials

was studied by Bonnell and Nottage [14] in England. Their

tests were performed on a material with a very low tensile

strength, saturated with a sodium sulfate solution. They

showed that ‘‘when hydration takes place within the pores,

A. Neville / Cement and Concrete R1278

stresses sufficiently high to overcome the tensile strength of

normal porous building materials may be developed’’ and

the salt may ‘‘exert a sufficient force to bring about

disintegration of the material’’ [14].

These were experiments involving thenardite and mir-

abilite (although these names were not mentioned),

although it was not concrete that was tested. The

experiments are highly relevant to the consideration of

physical sulfate attack on concrete, but the paper by

Bonnell and Nottage [14] is probably not well known

because it was published shortly before the outbreak of

World War II.

In 1949, Lea and Nurse [15] studied the crystallization of

solids in aqueous systems, using both sodium and calcium

sulfates. They identified a lack of knowledge of the growth

of crystals under stress, but, with respect to ettringite, they

favored the notion of topochemical reaction.

Consideration of through-solution as against topochem-

ical reactions is a matter for chemists. The proposition of

Lafuma [13] of topochemical formation is criticized by

Brown and Taylor [16] and also by Mehta [17].

The specific behavior of sodium sulfate salts has

continued to be a subject of study. Thenardite has a much

higher solubility than does mirabilite so that thenardite can

produce a supersaturated solution with respect to mirabi-

lite. In 2002, Flatt [18] expressed the view that at 20 jC, atensile hoop stress of 10 to 20 MPa develops by the

crystallization of mirabilite (decahydrate) from a saturated

solution of thenardite (anhydrite). Tensile stress of such

magnitude would inevitably disrupt concrete, but the pore

size influences the ease with which the crystals can grow

[18].

According to Flatt [18], ‘‘the development of crystalli-

zation pressure requires supersaturation in the liquid film

between the crystals and the pore wall’’, but he notes that in

concrete, ‘‘unfilled large pores tend to consume supersatu-

ration in harmless growth’’. Flatt [18] expresses the view

that ‘‘sodium sulfate can cause more damage than just about

any other salt’’.

In a 2002 paper, Brown [19] recognizes the distinction

between what he calls ‘‘classic form of sulfate attack

associated with ettringite and/or gypsum formation’’ and

‘‘physical sulfate attack associated with crystallization of

sulfate’’. He also acknowledges the fact that the involve-

ment of a sulfate is almost incidental to what is essentially a

physical attack by saying: ‘‘Physical sulfate attack can be

regarded a specific type of salt damage’’ [19].

The various papers that I have just discussed are

concerned with physical actions, the fact that sodium sulfate

has sulfate ions being incidental, so that they point toward

the concept of physical attack.

Overall then, I think it is useful to distinguish attack that

involves sulfate ions, i.e., sulfate attack, from a situation

where damage is caused solely by physical forces, which is

physical sulfate attack. I venture to recommend such

nomenclature.

1.6. How widespread is sulfate attack?

In a paper reviewing the progress in the practical field of

concrete in the last 40 years [20], I commented that in some

cases, the volume of laboratory research is disproportionate

to the practical extent of the problem. I cited the alkali–

aggregate reaction, but now, I wonder whether this is not

also the case with sulfate attack.

The opinion that sulfate attack is not a widespread

problem in concrete structures is expressed even by those

who have published extensively on the topic. Specifically, a

book titled Sulfate Attack on Concrete [6] says ‘‘that neither

external nor internal forms of sulfate attack are major causes

of degradation of concrete structures,’’ and, also, that

‘‘sulfate-related deterioration is globally minute’’ compared

with damage by corrosion of steel or by freezing and

thawing. The same book says—and I agree—that owners

of particular structures that have suffered sulfate-induced

damage are justified in being concerned. On the other hand,

the book contains a number of sweeping statements, ele-

vating the mere presence of sulfates to damage [6], sup-

ported by global references such as Deposition Transcripts

1996–2000. I am aware of some of these transcripts, and I

find in them also diametrically opposed views, opinions,

and interpretation.

Moreover, despite numerous allegations about damaged

foundations in California, not a single collapse has been

reported, nor has any house or building been evacuated.

In 1993, Mehta [21], who has repeatedly reviewed the

problems of sulfate attack, expressed the view ‘‘that sulfate

attack is seldom the sole phenomenon responsible for

concrete deterioration’’. Seven years later, he wrote that

‘‘the threat of structural failures due to sulfate attack. . .seems

to be even less of a threat than that caused by alkali–silica

reaction’’ [8].

At the 1998 Seminar on Sulfate Attack Mechanisms held

in Quebec City, Pierce [22], who was head of the U.S.

Bureau of Reclamation, said that he echoed Haynes’ words,

‘‘we do not see problems in the field’’, and ascribed this

situation to the fact that ‘‘we have taken the necessary

precautions’’. The precautions are incorporated in codes of

practice. Literature search indicates that in temperate cli-

mates, there have been very few, if any, actual failures.

The view that there are very few reported cases of

structures in service damaged by sulfates in the soil or in

groundwater is supported by Tobin [23]. In a technical

information note dated May 22, 2002, he refers to his 66

years’ experience with residential and commercial construc-

tion using concrete, and says, ‘‘Surface damages have been

observed, not unlike those that occur due to freeze–thaw

weather conditions on exposed concrete, but not a single

instance of a progressive attack or a complete failure due to

sulfate attack in the true meaning of the word’’ [23]. Those

are the words of a structural engineer.

These views contrast, to a large extent, with the

opinions of laboratory scientists who use accelerated test

esearch 34 (2004) 1275–1296

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1279

methods. Often, their objective is to study sulfate attack,

so that they have little interest in concrete that performs

well. I am not blaming them, but those dealing with real

concrete in service should not be swayed by tales of

doom.

There are, nevertheless, some parts of the world where

sulfates are problematic. In South Australia, especially in

the city of Adelaide, there is an endemic problem known as

salt damp. Essentially, the ground contains sulfate-laden

salt; this rises through the concrete, or other building

material such as limestone or rubble, in house foundations

during cool, wet winters. When the water evaporates during

hot, dry summers, salt is deposited on the surface of the

concrete. In consequence, concrete may be damaged. Pre-

vention in existing buildings made of porous materials is

generally uneconomic, and, usually, a sacrificial render is

applied and renewed periodically.

Parts of the Canadian Prairies abound in sulfates, and so

do some other areas in several countries. In the U.K.,

sulfates are found in some clayey soils and on so-called

brownfield sites, i.e., locations of former factories involving

chemical processes; these are not natural deposits.

1.7. Mechanism of sulfate attack

First of all, it is useful to list the principal reactions of

sulfates with hydrated cement paste. Those of main interest

are as follows.

Sodium sulfate reacts with calcium hydroxide to form

calcium sulfate (gypsum). This reaction proceeds to a

greater or lesser extent, depending on the conditions. In

flowing water, with a constant supply of sodium sulfate and

removal of calcium hydroxide, the reaction may eventually

continue to completion, i.e., leaching of all calcium hydrox-

ide (which, on a volume basis, is a major product of

hydration of portland cement). Otherwise, equilibrium is

reached.

Calcium sulfate formed as described above can subse-

quently react with C3A, usually via the formation of mono-

sulfoaluminate, to form ettringite.

Sodium sulfate is the predominant salt involved in

reactions with concrete in California.

Calcium sulfate reacts with C3A to form ettringite and

reacts also with sodium and potassium hydroxides. It is to

minimize the reaction with C3A that Type V cement was

developed.

Calcium sulfate is the predominant salt involved in

reactions with concrete in England.

Magnesium sulfate reacts with all products of hydration

of cement; the important resulting compounds are calcium

sulfate and magnesia. Calcium sulfate can proceed to react

with C3A.

The reactions of sulfates are well known so that there is

no need to write here the relevant equations in stoichiomet-

ric form. However, it is important to recognize that the end

products of the various reactions, if they occur so as to

damage the concrete, result in different types of damage. In

consequence, different preventive measures are required.

Specifically, in the case of sodium sulfate, the use of

cement with a low C3A content (which becomes calcium

aluminate hydrate) minimizes the extent of reaction with

sodium sulfate and, therefore, of the formation of ettringite.

This may lead to expansion, but, as pointed out by Taylor

[24], ‘‘seeing ettringite in a concrete does not necessarily

mean that ettringite has caused the stress’’.

As reported by Santhanam et al. [25], gypsum is the

primary product of sulfate attack at high concentrations of

the sulfate ion. When the pH of the pore water in hydrated

cement paste falls below about 11.5, ettringite is not stable

and decomposes to gypsum.

The formation of gypsum and its effects are still far from

clear. As recently as 2000, tests on pure C3S paste (i.e., in

the absence of C3A) showed that gypsum formation ‘‘may

be a cause leading to expansion and subsequent cracking’’

[26], which occurred after about one year in a 5% sodium

sulfate solution. The words ‘‘may be’’ are important. The

paper concluded that ‘‘it is not clear that gypsum formation

follows either the topochemical reaction mechanism or the

through-solution mechanism’’ [26]. The final words ‘‘more

research is needed in this area’’ [26] have a familiar ring.

Indeed, Mather [27] said, ‘‘I have seen no evidence that

the formation of gypsum during sulfate attack on cement

paste causes expansion’’. He favors the notion of a through-

solution formation of gypsum [27]. Harrison [5] expresses a

similar view.

In tests by Rasheeduzzafar et al. [28] on neat cement

paste and mortar specimens immersed in a solution of

sodium and magnesium sulfates, damage manifested itself

primarily by scaling, spalling, and softening, rather than by

expansion or cracking; there was a large reduction in

compressive strength [28]. Because magnesium hydroxide

has a very low solubility and its saturated solution has a pH

of about 10.5, C-S-H decomposes, liberating lime. This lime

reacts with magnesium sulfate and forms further magnesium

hydroxide and gypsum. This reaction continues until gyp-

sum is crystallized out [28]. Magnesium hydroxide reacts

also with silicate hydrate arising from the decomposition of

C-S-H, and this results in the formation of magnesium

silicate hydrate, which lacks cohesive properties [28]. These

reactions take place even when the C3A content in the

cement is very low [28].

Thus, it can be seen that magnesium sulfate is more

aggressive than sodium sulfate due to the lowering of pH of

the pore water in the hydrated cement paste. It follows that a

laboratory test using magnesium sulfate is not valid for field

conditions where the pH is not lowered during the sulfate

attack [21].

In this connection, I have to admit that my own labora-

tory tests [29] on sulfate attack on concrete were performed

using magnesium sulfate because, when I started the tests in

1965, I did not realize the peculiar consequences of that salt.

I also used a very high sulfate cement in the attacking

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961280

solution (50,000 ppm) as against several thousand parts per

million encountered in the field. The error of my ways

continues to be repeated because, as Hime [2] says, the

reaction is faster, which suits PhD students, and the fact that

it is different from that of calcium sulfate seems to elude

many investigators.

A further complication should be mentioned. Magnesium

hydroxide, called brucite, forms a protective layer on

concrete, unless this is damaged mechanically [30]. This

explains the good record of concrete in seawater, which has

a high magnesium sulfate content.

It might be thought by some readers that the preceding

apparent literature review is too academic for a paper

purporting to deal with sulfate attack in service. My

defense is that the seeming lack of clarity allows one-

sided, and therefore biased, views on the lines that ‘‘if

gypsum is present, sulfate attack has occurred’’ or ‘‘if no

ettringite has been found, there is no sulfate attack’’.

Moreover, often, experience with, say, magnesium sulfate

is directly transferred to a situation involving sodium

sulfate. Conversely, laboratory test results on a single type

of sulfate are applied to field situations, in which a mixture

of sulfates is present.

With reference to exposure conditions, it is worth noting

that the carbonation of a concrete surface prior to exposure

to sulfates ‘‘reduces the build-up of sulfate within concrete’’

[5].

2. Part II: Progress of sulfate attack and preventive

measures

This part of the four-part paper discusses the progress

and consequences of sulfate attack, mainly under field

conditions, and also some preventive measures.

2.1. Progress of attack

It is obvious that sulfate attack can occur only when

sulfates are present in contact with concrete. But how much

sulfate needs to be present for the reactions between the

sulfates and the hydrated cement paste to result in damage to

concrete, i.e., to constitute sulfate attack? We should also

ask the question whether the concentration of sulfate in the

water in contact with concrete affects the extent of the

attack; the extent comprises the intensity and the rate. These

depend on the solubility of the given sulfate and on the

continuity, or otherwise, of the supply, i.e., on whether the

water is static or is flowing.

The preceding paragraph could be interpreted to refer

first to sulfates in the soil, but then to continue with

consideration of sulfates in water. The explanation of this

apparent inconsistency is that dry salts do not react with

concrete so that it is the concentration of sulfates in water in

contact with concrete that is relevant. In other words,

sulfates can move into concrete only in solution.

The rate of attack by sulfates is affected by the strength

of the sulfate solution, but the effect becomes small beyond

about 1% of sodium sulfate and about 0.5% for magnesium

sulfate [31]. This effect is of significance in the classifica-

tion of aggressivity of exposure, given in the various codes,

and is discussed in a later section.

Taylor [32] lists the following as consequences of sulfate

attack: strength loss, expansion, cracking, and, ultimately,

disintegration. He further points out that whereas laboratory

studies have concentrated on ‘‘expansion and cracking, field

experience shows that loss of adhesion and strength are

usually more important’’ [32]. This accords with my criti-

cism of excessive reliance on individual, and often discon-

nected, laboratory studies coupled with an absence of in-

depth (no pun intended) evaluation of field experience.

Similarly, Brown and Taylor [16] say, ‘‘Further experimen-

tal work should take into account the conditions existing in

the field and be related to field observations’’.

As for changes in concrete resulting from the action of

sulfates, Skalny et al. [6] list nine chemical changes and

nine physical changes, but say that ‘‘by themselves, neither

(sic) of the above chemical, physical, and microstructural

changes are necessarily an adequate sign of sulfate attack’’.

They say further: ‘‘However, in combination, there can be

little doubt’’ [6]. This is a very sweeping statement because

much depends on which combination is involved.

References to the action of sulfates in Southern Cal-

ifornia should not be generalized because, as stated by

Skalny et al. [6], soils there contain ‘‘high levels of

sulfates, often in the form of gypsum’’. As I have already

pointed out, the action of calcium sulfate and its conse-

quences are different from the action of sodium and

magnesium sulfates. Thus, unqualified statements, such

as ‘‘one of the characteristic features of severe external

sulfate attack is formation of gypsum’’ [6], are too vague

and may be misleading to engineers without specialized

knowledge of chemistry.

2.2. How to make sulfate-resisting concrete

It would be contrary to the tenor of this paper if I were to

make a categorical statement of the ‘‘one-size-fits-all’’ type

about how to ensure sulfate resistance of concrete. However,

I have come across forceful assertions that for concrete to be

resistant to sulfates, all that matters is a low w/c. An

experienced engineer knows that unless concrete is well

compacted and dense, a low w/c is useless. DePuy [1] gave

as a first requirement for good resistance, ‘‘a high-quality

impermeable concrete. This requires good workmanship

(workable mixes, good consolidation, a hard finish, and

good curing), and the use of a relatively rich mix with a low

w/c.’’ Most engineers would subscribe to this view, but

some laboratory scientists might have little experience of

these site-related factors.

I shall deal with w/c in a later section but, by way of

introduction to the concept of sulfate-resisting concrete, I

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1281

have to mention also the use of Type V, i.e., sulfate-resisting

cement. This will be considered in the next section.

At this stage, I would suggest that there is no single

answer to the question often posed to me: Which is more

important, w/c or the type of cement? The main reason for

the absence of a unique answer is that the mechanism of

sulfate attack depends on the cation in the sulfate: sodium,

calcium, or magnesium. Of course, more than one of these

may be present in a single location. Alas, in disputes, an

expert may be expected to present a single answer.

Parenthetically, I would like to refer to another problem

occasionally faced by an expert. He or she is asked whether

it is certain that, under some circumstances, damage cannot

occur. Alas, lawyers cannot, or do not want to, accept that a

certainty of impossibility of the occurrence of an event does

not exist in engineering.

Apropos of making a categorical statement that is uni-

versally right, Northcote Parkinson, a 20th century English

satirical writer, commented that no one is always right, but

that in every organization, there is one person who is

invariably wrong. That person is very useful as a sounding

board for a new idea claimed to be a winner: if that person

thinks so too, then, the idea is wrong!

2.3. Use of sulfate-resisting cement

This type of cement (known as Type V) has existed in the

United States since 1940, although since then, ASTM has

introduced some modifications in the standard requirements

for its chemical composition.

The essential feature of Type V cement is that it limits the

content of C3A, as calculated by the Bogue method.

However, as pointed out by Mehta [21], Type V cement

‘‘addresses only the problem of sulfate expansion associated

with the ettringite formation’’. I would expect, therefore,

Type V cement to be particularly efficacious when calcium

sulfate is the attacking medium, although it could be

beneficial with respect to the prevention of the formation

of gypsum owing to the action of sodium sulfate.

Thus, Type V cement is of no avail in the attack of

calcium hydroxide and C-S-H and the subsequent loss of

strength [21]. This situation is relevant to tests on sulfate

resistance: do we determine expansion or loss of strength?

These tests are discussed in Part III of this paper.

We should note that for exposure to modest amounts of

sulfates, codes allow the use of Type II cement with

pozzolans as an alternative to Type V cement.

Nevertheless, the importance of using Type V cement

should not be underestimated. The Bureau of Reclamation

Concrete Manual stated, in 1985, that ‘‘concrete containing

cement with a low content of the vulnerable calcium

aluminate is highly resistant to attack by sulfate-laden soils

and water’’ [33]. Even today, the Bureau of Reclamation is

reported to consider ‘‘the C3A content to be the greatest

single factor influencing the resistance of portland cement

concrete to sulfate attack, with cement content second’’

[31]. Some specific values of the C3A content in cement

and of the cement content in concrete are quoted, but I

suspect that it is the density of concrete and not the cement

content as such that is the real factor. The view that cement

content has no effect on sulfate resistance was expressed

also by The Concrete Society [34] in 1999.

Recent interpretations of long-term tests at the Bureau of

Reclamation will be discussed in a subsequent section.

Although dated, the views of Lea [35] are of importance.

He stated, ‘‘The resistance of Portland cements to attack by

sulphate solutions is known to be related in a general way to

their calculated C3A content, but there are many anomalies

in this relationship’’ [35]. One reason is that the rate of

cooling the clinker influences the morphology of the alu-

minate: if it is in glass form (amorphous), it is much less

susceptible to attack by sodium or magnesium sulfate [35].

Also somewhat dated, but of interest, are the views of

Mather [36]. He said that when the Bureau of Reclamation

recommendations are observed, i.e., Type V cement is used,

‘‘no significant deterioration of concrete due to sulphate

attack has been encountered’’ [36].

An enormous contribution in the field of sulfate attack on

concrete was made by Thorvaldson in the 1950s [37]. (In

1962, I had the privilege of knowing him, and I profited

from his wealth of knowledge of cement.) He confirmed the

influence of the type of cement on the resistance to sulfate

attack of concrete with a high w/c. As usual, the studies

were comparative with respect to Type I cement. However,

Type I cement has changed considerably since then, mainly

by way of an increase in the ratio of C3S to C2S. In

consequence, the products of hydration contain more calci-

um hydroxide than is generated by C2S. This makes the

cement more vulnerable to sulfate attack, with the conse-

quence of the formation of gypsum. Al-Amoudi [38] ob-

served that in the case of magnesium sulfate, a lower content

of calcium hydroxide in hydrated cement is undesirable

because it encourages the reaction of sulfate with C-S-H,

leading to the softening of the matrix, mass loss, and a

reduction in strength.

The positive influence of a low C3S-to-C2S ratio was

reported also by Irassar et al. [39]. In my view, their paper is

an example of an investigation involving a single set of

conditions, and great care is required before generalizations

can be made. At least on evidence to date, we should refrain

from saying that a low C3S-to-C2S ratio is a significant

positive factor in the choice of cement for good sulfate

resistance.

Furthermore, consideration of the ratio of C3S to C2S is

often not included in the interpretation of laboratory tests. In

the case of concrete in situ, the ratio is usually unknown.

Nowadays, in addition to portland cement, there are

commonly used various cementitious materials, such as

fly ash and ground granulated blastfurnace slag (slag). They

inevitably affect the chemical reactions involving sulfates. I

am not considering these cementitious materials because the

situation when only portland cement is used should be

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961282

disentangled first. With additional cementitious materials,

the complexity of sulfate attack becomes even greater. For

example, Taylor [24] pointed out that if slag has a low

alumina content, it improves the sulfate resistance, but with

a high content of alumina, the reverse is the case. On the

other hand, if calcium hydroxide becomes bound by, say, fly

ash, the sulfate resistance may improve.

A further complication arises from a possible interaction

between the type of cement and the w/c. This makes it

difficult to interpret test results in which one parameter at a

time is varied. To illustrate the situation, I shall refer to the

works of Stark. He conducted long-term studies and

reported his findings in 1989 [40] and again in 2002 [41].

According to the first report of Stark [40], visual assess-

ment of concrete specimens half-submerged in a sodium-

sulfate-rich soil, with periodic wetting and drying, showed

that Type V cement performed better than did Type I cement

at values of w/c between 0.75 and 0.45. At w/c = 0.36, the

two cements were equally good. To me, these results show a

superior performance of Type V cement and not a predom-

inant influence of w/c. A generalization is, however, not

possible. It is worth noting that Stark [40] commented that

‘‘wetting and drying with crystallization of salts may have

contributed to deterioration’’.

The second report of Stark [41], dated 2002, also relies on

visual assessment of deterioration. He found that at a w/c of

0.40, concrete, partly immersed in sulfate-laden soil and

alternately wetted and dried, performed significantly worse

when Type I cement was used than with Type V or Type

II cement [41]. The w/c had a significant effect on the

performance above a w/c of 0.50. At a w/c of 0.65, all ce-

ments led to a poor performance but, at intermediate values

of w/c, Stark [41] found the cement composition to be ‘‘of

some importance’’. Interestingly, laboratory concrete im-

mersed continuously in a solution of sodium sulfate of the

same concentration as in the field tests (65,000 ppm of

sulfate) showed uniformly excellent performance regardless

of the type of cement, w/c, or additional cementitious

materials and surface treatments [41]. The conclusion of

Stark [41] is that ‘‘the traditional explanation of expansion

of concrete due to chemical reaction of sulfate ion with

aluminate-bearing cement hydration products was of rela-

tively minor significance’’. He concluded that a major

mechanism of deterioration in the field tests (but absent in

the laboratory) is cooling and heating, and wetting and

drying (by natural variation) with cyclic crystallization of

sodium sulfate salts [41].

According to the classification that I have used in this

paper, the above mechanism should be considered to be

physical attack. Much more importantly, the tests of Stark

[41] show that predicting the behavior in the field on the

basis of laboratory tests under singular, uniform, and strictly

controlled conditions is likely to be unreliable. This explains

why various laboratory tests lead to different conclusions.

Hence, my plea for a new approach to the study of sulfate

attack is by way of field performance.

Actually, varying one parameter at a time in a concrete

mix is not possible. For example, at a given workability, a

change in the w/c entails a change in the cement content.

2.4. Importance of w/c

Occasionally, especially in disputes, strong emphasis is

placed on the low value of w/c being the essential require-

ment for sulfate resistance, the type of cement being

considered irrelevant or, at best, secondary. Other people

express the contrary view. Each proponent adduces evidence

supporting his or her opinion. Such a situation is highly

conducive to the continuation of the present confusion, but

it is not surprising.

It is not surprising because much depends on the cation

in the attacking sulfate, as shown earlier in this paper. There

are other important factors at play as well.

There is, however, an overriding requirement, necessary

but not sufficient, for good sulfate resistance of concrete,

and that is its density in the sense of a very low permeabil-

ity. This is achieved by good proportioning of the mix

ingredients, such that full compaction can be obtained, by

actual achievement of full compaction, and by effective

curing to maximize the degree of hydration of the cement

paste. These requirements are obvious to an engineer, but

often not fully considered by laboratory experimenters.

Hence, my repeated plea for consideration of field concrete.

Building Research Establishment (BRE) Special Digest 1

states: ‘‘Poorly compacted concrete will be particularly

vulnerable to chemical attack’’ [42]. Taylor [32] says that

‘‘the effects (of sulfate) are minimized in a dense concrete of

low permeability and by using a sulphate-resisting cement,

in which there is little or no aluminate phase. . .’’.Consideration of low permeability to ensure durability is

included in British Standard BS 5328, which says that

permeability is ‘‘governed by the constituents, their propor-

tions and the procedures used in making concrete’’ [43].

Eight factors influencing durability are listed, of which one

is the type of cement, and another the cement content and

free w/c taken together [43].

Permeability was also considered in the tests of Khatri et

al. [44], who found that the expansion of mortar specimens

immersed in a 5% sodium sulfate solution was affected both

by the type of binder and by permeability.

With reference to the influence of w/c on the resistance of

concrete to external attack in general, it is useful to quote

Shah et al. [45]: ‘‘a relatively low w/c in cement paste does

not necessarily mean a reduced permeability because of the

significant influence of aggregates’’. Specifically, they con-

sider early-age cracking to be a factor in ensuring durability.

The relevance to the theme of this paper is that absolutist

statements about w/c being the governing factor ensuring

resistance to sulfate attack are not correct.

Young et al. [46] say that sulfate attack can be prevented

by any one of three factors: a low w/c, a low calcium hyd-

roxide content, or a low C3A content. If their claim is valid,

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1283

then adequate protection of concrete should be ensured by

the use of a low w/c alone or by the use of Type V cement

alone. In my opinion, these solutions are too sweeping and

not valid under all circumstances.

Somewhat surprisingly, it has been found in several

investigations, as reported by Al-Amoudi [38], that lower-

ing w/c has a deleterious effect on the resistance of concrete

exposed to magnesium sulfate, the use of Type V cement

being of no avail. A likely explanation is that at low values

of w/c, there is limited pore space to accommodate the

products of reactions with sulfate, namely, magnesium

silicate hydrate (which has no adhesive properties) and

gypsum. Care is required in translating these observations

into practical recommendations about mix selection. My

purpose in reporting the above findings is to show, once

again, that a systematic review of the entire problem of

reactions of sulfates with hydrated cement, for the purpose

of guidance about making durable concrete, is still to come.

Specifically, we often fail to differentiate between the

cations in guidance documents or codes, or when applying

such documents.

2.5. Bureau of Reclamation tests

The Bureau has played a preeminent early role in

studying sulfate attack on concrete and in developing

sulfate-resisting concrete. An extensive and independent

long-term investigation of the behavior of concrete exposed

to sodium sulfate was started by the Bureau of Reclamation

in the 1950s. The specimens were continuously exposed to

sulfate over a period of more than 40 years. In addition, the

Bureau conducted accelerated tests. Unfortunately, because

of funding changes, the Bureau did not analyze and report

the long-term test results.

The spate of litigation involving alleged sulfate attack on

concrete led several people to visit the Bureau test results

and to publish their interpretations. The steady-exposure,

i.e., nonaccelerated tests, were analyzed by Kurtis et al. [47].

It is not my intention to comment in detail on their analysis,

but I believe the following observations are in order. First,

the extensive development of what they call ‘‘empirical

model to predict concrete expansion’’ in their paper was

revoked by an erratum published eight months later [47],

and equations with new shapes were introduced. This is a

great pity. Less hurried research would serve us better.

Another aspect of the interpretation by Kurtis et al. [47]

of the test data of the Bureau of Reclamation deserves

mention. The interpretation distinguishes cements with a

low C3A content (described as < 8%) and cements with a

high C3A content (described as > 10%). This gives the

impression that the Bureau tested cements with quite a range

of C3A contents. It is true that the range was wide, but the

distribution of C3A contents was uneven, and the division at

10% is fortuitous: there were no cements with a C3A content

between 8.0 and 10.0. Moreover, the distribution of values

was not uniform: 81 cements with less than 8.0% and 33

with more than 10.0%; of these 33 cements, 16 had a C3A

content of between 16.01% and 17%. In addition, of the 81

cements with the C3A content of less than 8.0%, 43 had a

content higher than 6.01% and 53 higher than 5.01%. The

relevance of the value of 5.01% will be mentioned in the

next paragraph.

My remarks are not meant to criticize the findings of

Kurtis et al. [47] because it was not they who chose the

cements (and the Bureau of Reclamation had its own

objectives), but their interpretation in qualitative terms, such

as ‘‘low C3A content’’ and ‘‘high C3A content’’, may not be

justified. Even more importantly, we should note that

ASTM C 150-2002 limits the C3A content of Type V

cement to 5% and also restricts the sum of the content of

C4AF plus twice the C3A content to 25%. Thus, as I see it,

conclusions in terms of high- and low-C3A contents need

careful qualification.

I have a similar reservation with respect to the use of w/c

as a parameter. Whereas the value of w/c used by the Bureau

ranged from 0.35 to 0.75, of the 114 mixes used, 85 had w/c

between 0.46 and 0.55. Thus, the real spread of w/c was

minimal. This is relevant to the conclusions about the

significance of w/c with respect to the sulfate resistance of

concrete.

Monteiro and Kurtis [48] (who were the coauthors of Ref.

[47]) wrote another paper discussing the same test results.

Although they emphasize the importance of the influence of

w/c on the sulfate resistance of concrete, this is not entirely

borne out by the test results. Their conclusion that the ‘‘time

to failure, as measured by expansion, decreases with in-

creasing w/c and C3A content’’ is correct, but it is far too

broad for practical application. At one extreme, they report

that the concrete that failed first (after 3½ years) had a w/c of

0.48 [48]. This concrete was made with a cement that

contained as much as 73.7% of C3S; hence, the C3S content

may be significant. At the other extreme, of the nine

concretes that failed in less than 25 years and had a w/c

>0.55, seven had a C3A content above 6%. Moreover, a

graphical interpretation of the tests, reported by Monteiro

and Kurtis [48], shows that for values of w/c between 0.45

and 0.55, the time to failure ranged from 3½ years (but

mostly >12 years) up to more than 40 years. My view of

their analysis is that it does not enable us to draw any

practical conclusions, especially because only the expansion

of specimens was used as a measure of sulfate attack and

only exposure to sodium sulfate was tested.

In view of this, it is surprising that they report [48] a

reliability analysis (of which they were coauthors) that is

said to have shown ‘‘that the effect of w/c on expansion was

one order of magnitude higher than the effect of C3A

content for cements with a C3A content of less than 8%’’

[49]. That phrase is followed by the following parenthetical

qualification: ‘‘for cements with low C3A’’. They mean less

than 8% and not a Type V cement, whose upper limit is 5%.

Strangely enough, I cannot find the exact wording attributed

to Corr et al. in Ref. [49].

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961284

It could be thought that a person who uses (or misuses)

the Bureau of Reclamation test data to prove that w/c is the

sole factor governing the sulfate resistance of concrete

‘‘doth protest too much’’.

The test results of the accelerated tests conducted by the

Bureau of Reclamation, starting in the early 1950s over a

period of 18 years, were analyzed by Ficcadenti [50]. The

tests used soaking the concrete specimens alternately in a

sodium sulfate solution and drying in hot air; the expansion

of the specimens was measured. What is significant is his

finding that the expansion does not progress linearly but

slows down or ceases when C3A has been used up.

Ficcadenti [50], therefore, concluded that the C3A content

of the cement is the crucial factor, and w/c has little

influence on expansion over long periods.

I do not have enough information to try to reconcile the

findings of Ficcadenti [50] with those of Kurtis et al. [47] or,

alternatively, to deduce the correct interpretation of the

Bureau of Reclamation tests, nor is such a resolution of

conflicting conclusions within the ambit of this paper. My

purpose in mentioning the varying interpretations of the

work of the Bureau of Reclamation is to illustrate the

confused and confusing state of our knowledge of sulfate

attack on concrete. Possibly, someone will undertake an

independent retrospective study of all of the Bureau test

results.

3. Part III: Assessment of sulfate attack

In this part, I discuss the approach of various codes to the

classification of the severity of exposure conditions prior to

construction. This is followed by consideration of tests to

determine the damage to concrete in situ. Appendix A of

this paper contains selected tables from various sources

giving a classification of exposure to sulfates.

3.1. Severity of exposure to sulfates

The preceding discussion of sulfate attack was in qual-

itative terms, but for practical purposes, we need to relate

the degree of severity of field exposure to the extent of the

attack on concrete. In the American and British standards

(until 2003), there exist two methods of characterization of

the exposure: sulfate content in the ground water that is

likely to be in contact with the concrete, and the sulfate

content in the soil that will be in contact with concrete. Two

general comments are pertinent.

First, how comparable are the severities of exposure

when determined on water and on soil? This question will

be dealt with in a later section, and so will the related issue

of the extraction ratio from the soil.

Second, in all cases, the quality of sulfate is expressed

in parts of SO4 in milligrams per kilogram of water, i.e., in

ppm (parts per million), generally without consideration of

the type of cation in the sulfate. And yet, we have seen

that different cations may result in different forms of

attack. Incidentally, some earlier standards expressed the

sulfate as SO3. While it is easy to convert the amount of

SO3 into SO4, care is required when using different

documents.

The British Code of Practice for structural use of concrete,

BS 8110, 1985 [51], recognizes four classes of exposure,

plus a class (less that 300 ppm in groundwater or 1000 ppm

in a 2:1 water–soil extract) that has no requirements with

respect to the concrete mix. In 1997, there was published

British Standard BS 5328 on the selection of mix proportions

[43]. This standard recognizes magnesium sulfate as a

separate criterion of severity of exposure.

BS 5328 has now been withdrawn, and the current

British standard is the European Standard EN 206-1 [52].

With respect to sulfates, expressed as SO4, three classes of

severity are recognized (the harmless condition not being

considered). It is interesting to note that the classification

by exposure classes and the limiting values in EN 206-1

are not recommended for use in the U.K. Instead, a

complementary British Standard to EN 206-1, known as

BS 8500-1, has been published. One shortcoming of EN

206-1 is that it uses the total sulfate in the soil (obtained by

hydrochloric acid extraction) but allows water extraction if

local experience is available. In the U.K., between 1939

and 1968, the sulfate content in the soil used to be

determined by acid extraction (although this was not

explicitly stated), and this greatly overestimated the sever-

ity of exposure.

In the United States, ACI Building Code 318 [53] re-

cognizes three classes of exposure in terms of SO4 (plus the

harmless condition) and prescribes a maximum value of the

water/cementitious material ratio, the minimum compres-

sive strength, and the type of cement. The latter is selected

on the basis of the C3A content. The Commentary on ACI

318 lists ‘‘other requirements for durable concrete exposed

to concentrations of sulfate’’; these include adequate com-

paction, uniformity, and sufficient moist curing [53]. The

importance of compaction and curing was emphasized

earlier in this paper.

In addition to design codes and standards, in the U.K.,

there exist so-called Digests, which are nonmandatory

guidance documents, perhaps like the ACI guides. Guides

relating to the sulfate resistance of concrete have existed for

many years, but were revised from time to time. A selection

of extracts from various American, British, European, and

Canadian sources giving a classification of the severity of

exposure to sulfates is given in Appendix A.

In 2002, there was published a new BRE Special Digest

1 [42], dealing with concrete in aggressive ground. It is a

long and complex four-part document, which provides

design guidance for various types of buildings. Part 1

introduces six classes of aggressive chemical environment

in terms of soluble sulfate, magnesium, potential sulfate

from oxidation of pyrite (not considered in the present

paper), pH, and mobility of ground water [54].

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1285

The class of aggressive chemical environment is com-

bined with the ‘‘structural performance level’’, which is

categorized according to the service life, criticality of use,

and structural details. This useful innovation distinguishes

three levels: (i) structures expected to have a life of less than

30 years and unreinforced house foundations; (ii) structures

with an expected life of 30 to 100 years; and (iii) structures

expected to have a life in excess of 100 years and critical

parts of structures, such as slender elements and hinges.

This approach represents a great improvement on the

traditional approach in existing codes, which uses a one-

size-fits-all approach.

The combination of the class of aggressive chemical

environment with the structural performance level deter-

mines the so-called design class. There are 13 design

classes, each of which can be combined with three structural

performance levels, and three cross-section thicknesses.

Hence, the number of ‘‘additional protective measures’’

required is obtained; these include surface protection, a

sacrificial layer, and use of controlled permeability form-

work. Finally, this information leads to mixes specified in

terms of type of cement (or cement group, in European

parlance), minimum cement content, and maximum w/c.

It is useful to note that the size of the cross-section of the

concrete member is taken into consideration in determining

the structural performance level. This means that this

approach implicitly accepts that some attack on the surface

of concrete can be tolerated in thicker sections [54]. How-

ever, fine-tuning according to the thickness of individual

members on a single site is, in my opinion, impractical in

terms of mix delivery on site.

Consideration of mobility of ground water in an explicit

manner, rather than casually, and advice on the determina-

tion of mobility of the water, introduced in Special Digest 1

[42], are a very useful innovation.

As I mentioned earlier, BRE Special Digest 1 [42] is not

a mandatory document but is intended to serve as a guide. It

is thus a competent source of advice for a competent

engineer, but it should not be used as a rulebook to be

followed, more or less, blindly.

Those concerned with investigations of allegedly dam-

aged concrete should note that the design of structures using

Part 3 of the Special Digest 1 ‘‘is not intended as a basis for

assessing the risk of sulfate attack to existing properties’’

[42].

The system devised in the BRE Special Digest 1 [42] is

sound and admirable. Alas, in my view, it looks like a

product of an extensive desk study, with little consideration

of the realities on site, where, often, there does not exist a

single set of physical conditions. I shall discuss the vari-

ability of site conditions in a later section.

I realize that it is not easy to write a comprehensive

alternative set of recommendations for the choice of the

concrete mixes to ensure resistance to sulfates, but I suspect

that what is proposed in BRE Special Digest 1 is unlikely to

find favor with designers. The choice of a mix is on the

basis of 16 classes of aggressive environment, for three

structural performance levels, and three thicknesses of the

concrete member, compounded by ‘‘additional protective

measures’’ that raise or lower the quality of the concrete

mix. All this is a true embarras de richesse. Giving a choice

of all the 124 single malt whiskies at the same time might be

too much for a regular drinker.

What should be of greater concern are the class bound-

aries of the various exposure conditions. In a sense, they are

arbitrary because they have not been calibrated by the

measurement of recorded incidence of damage to concrete

caused by sulfate attack. This, of course, would not be easy

because of the variability of conditions in the field. Hence,

although we are not sure about the exact position of class

boundaries, the approach currently used is qualitatively

satisfactory. However, notice should be taken of the pre-

dominant cation in the sulfate because this is relevant to the

efficacy of using a Type V cement.

3.2. Extraction of sulfate from soil

The various codes classify the degree of exposure of

concrete to sulfates according to the sulfate content in

ground water or in an extract from the soil. Presumably, it

is intended that either of these determinations should result

in the same class of severity of exposure for the purpose of

mix selection.

The reason for the dual approach is that where a borehole

does not contain static or flowing ground water, it is

necessary to resort to testing the soil. Where there is a

choice, it is water that should be tested; this is explicitly

recommended by BRE Special Digest 1 [42].

Why then am I discussing the two test methods? Whereas

determining the sulfate content in water is a matter of a

simple chemical procedure, in the case of an extract, the

outcome of the same procedure will be affected by the

method of obtaining the extract. The method is by no means

standardized or uniform in various codes.

I should like to add that I am discussing the determina-

tion of the content of water-soluble, and not of acid-soluble,

sulfates. The latter method is used in some codes for certain

purposes, but I view it as inappropriate because the attack of

concrete, if any, occurs by water in contact with its surface

penetrating into the interior of the concrete. On the other

hand, acids may dissolve sulfates that have only a low

solubility in water.

The extract should be described by mass of water per unit

mass of soil; this varies from 1:1 to 20:1. I understand that

the California Department of Transportation [55] uses an

extraction ratio of 3:1, and the U.S. Bureau of Reclama-

tion and the U.S. Army Corps of Engineers use 5:1. The

significance of the extraction is that, under some circum-

stances, a high extraction ratio will give a higher sulfate

content, and this will lead the designer to choose a

concrete mix with superior properties than otherwise

would be the case.

Method A 7000 7000 10,000 9000

Method B 5000 11,000 8000 11,000

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961286

Many earlier standards were silent on the extraction ratio.

In consequence, and not surprisingly, when a party inter-

ested in demonstrating a high sulfate content was in control,

it used a high ratio, sometimes as high as 20:1, whereas the

opposing party would use a very low ratio, possibly down to

1:1. The recent Special Digest 1 prescribes a 2:1 extraction

ratio [42]. This has been used in the U.K. since 1975 and

replaces the 1:1 ratio used between 1968 and 1975. The

change was made because, with the 1:1 ratio, some clays

yielded an insufficient amount of extract water for chemical

tests; this problem, however, could be remedied by taking a

larger sample of soil.

The significance of the extraction ratio depends on the

type of sulfate prevalent in a given situation. Specifically, if

sodium and magnesium sulfates are dominant, and calcium

sulfate is a minor component, then a low extraction ratio

may be appropriate because both sodium and magnesium

sulfate have a high solubility, so that even a small amount of

water will result in an appropriate measure of sulfate

exposure. On the other hand, with calcium sulfate, a low

extraction ratio may lead to a measured value of sulfate

content corresponding to the maximum solubility of calcium

sulfate, whereas the actual sulfate content may be much

higher. This situation obtains when the sulfate content is

expressed in milligram of sulfate per kilogram of soil.

To give an example, assume that the content of calcium

sulfate in the soil is such that there are 10,000 mg of sulfate

per 1 kg of soil. Now, the low solubility of calcium sulfate

means that a maximum of 1440 of sulfate (SO4) can be

dissolved in 1 kg of water. Thus, a 1:1 extract would

indicate 1440 ppm, which would be reported as 1440 mg/

kg of soil, although the actual content is 10,000 mg/kg.

Now, using an extraction ratio of 20:1, the amount of

calcium sulfate would be 10,000 ppm or 10,000 mg/kg,

which is the actual value in the soil because all the sulfate in

the soil would be dissolved without reaching saturation of

the water.

I hesitate to offer an opinion on the choice of an

appropriate extraction ratio or on the class of boundaries

for the severity of exposure to sulfates. However, I find the

argument for expressing the limits in terms of sulfate in an

extract solution, and not as sulfate in the soil, to be

compelling. Thus, the European approach is preferable to

the ACI method.

I believe ACI is considering a change in the method of

expressing the sulfate content in soil. One proposal is to use

the term ‘‘sulfate activity index’’. This term is uninforma-

tive. Another proposal is to use the term ‘‘sulfate in pore

water of saturated soil’’. This is cumbersome and may

mislead some people who associate the term pore water

with water in the hydrated cement paste. Moreover, the soil

tested may not always have pores in it saturated, and the

extract may contain more water than could fill the pores in

the soil in situ. My support, for whatever it is worth, is for

‘‘sulfate in extract solution’’, with the extraction ratio

uniquely defined.

The difficulty of establishing the sulfate content of the

soil is illustrated by the data of Marchand et al. [56] who

compared the values obtained by extraction using the

California Transportation Procedure 417 (Method A; [55])

with tests on centrifuged liquid from the soil, followed by

ion chromatography, to yield the value of sulfate (Method

B). Marchand et al. [56] present a plot relating these two

sets of test results but make no statement about confidence

limits. By eye, there seems to be a large scatter; for example,

here are four pairs of results (in ppm of SO4):

Unless a statistically sound explanation of these results is

available, we should beware of making definite claims about

relationships. Rather, we should establish a reliable and

universally acceptable method of determining the sulfate

content of the soil.

Finally, there exist some newer chemical methods, such

as inductively coupled plasma atomic spectroscopy, and

they may not yield the same results as other methods, but

this is for chemists to say.

3.3. Taking soil samples

At the beginning of this paper, I admitted that even

though I am not a chemist, I felt justified in writing about

chemical attack on concrete. At this stage, I would like to

write about some soil aspects of sulfate attack on concrete,

even though I am not a geotechnical engineer. (Of course,

these topics were included in my civil engineering curricu-

lum.) Why? The answer is that both chemical and geotech-

nical considerations are provided for engineering purposes:

construction is preeminently an engineering matter.

In documents dealing with sulfate attack on concrete,

there is generally no guidance on taking soil samples. There

is a recent exception to this, namely, the BRE Special Digest

1 [42], which gives advice on taking samples and on the

interpretation of the test results on sulfates in water and in

the soil. Generally, the distribution of sulfates in soil varies

both horizontally and vertically. In the U.K., often, the top

meter or so of soil is relatively free from sulfates because of

leaching by groundwater [35]. On the other hand, in dry

regions, where the rate of evaporation is high, there may be

a higher concentration near the surface [35].

In my, admittedly limited, experience of large sites for

multiple buildings construction, I learned that there may be

a considerable variability in the different factors involved.

For example, if the site is sloping, the conditions of

groundwater may vary from place to place. They will vary

also depending on whether sampling was done just after a

heavy rainfall or after a prolonged period of drought. The

conditions can also vary on a given site in consequence of

excavation elsewhere. The combinations are numerous. And

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1287

yet, for practical purposes, the foundations for a number of,

say, houses cannot be finely tuned to the conditions estab-

lished at every individual site. Moreover, the measurements

may well be taken prior to exact siting of every individual

house.

At the same time, it would be prohibitively expensive to

go on testing until the worst spot has been found and to

make the conditions there govern the entire site. Never-

theless, such an extreme approach has been taken: it was

said that if one part of the land has a high sulfate content

and the rest only a low or even negligible content, then the

high content governs. The opinion expressed was: ‘‘take

enough results and you’ll find that it (high sulfate content)

exists throughout’’. In other words, all the foundations

have to be designed for the most severe condition encoun-

tered. Accepting such reasoning leads to demonstrably

nonsensical consequences.

Let us say that we propose to construct 100 individual

homes. Let us assume that there is an adjacent existing

development on a soil in which a high sulfate content was

established some time ago. Given that the soil and sulfates

are not aware of ownership boundaries, a high sulfate

content in the soil on an adjacent property means, according

to the logic quoted above, that all 100 proposed homes need

to have foundations that will resist the high sulfate content

on the adjacent property. It follows further that no testing for

sulfate is necessary. This saves money that would otherwise

be used for testing, but a great deal of money may be wasted

by an unnecessary provision for resistance to a high sulfate

content in the soil. Indulging in reductio ad absurdum, the

conditions on a single site could be extrapolated to be valid

further and further away.

Some help in solving this dilemma is offered by Special

Digest 1 [42], which recommends calculating a characteris-

tic value, rather than taking the maximum value, except

when only a small number of soil samples has been taken.

When there are five to nine results, the mean of the two

highest values is taken as the characteristic value. And when

there are more than 10 results, the mean of the highest 20%

is used [42].

This advice is helpful, but it is less clear what to do on

very large sites, such as those for, say, 100 individual

houses. Much depends on the uniformity of soil in geolog-

ical terms and on the extent of disturbance and grading that

the soil has been subjected to. I am not proffering advice

because this problem is more within the province of soil or

geotechnical engineers.

Because water is essential for sulfates to interact with

concrete, the possible movement of water should be con-

sidered in the construction of foundations. When these are

shallow, the isolation of concrete from the ambient material

is possible. Alternatively, or additionally, drainage around

the foundations can be provided. Similarly, a means of

carrying rainwater from the roof away from the foundations

should always be provided, even if the climate is thought to

be dry.

I am sure I am not alone in having observed a drainage

system vitiated by the house owner who extended the

garden right up to the house wall and then provided

continual watering so as to make the concrete almost

permanently wet. The situation can be aggravated by the

use of fertilizers for plants in close proximity to foundations.

Many fertilizers contain potassium and magnesium sul-

fate. I understand that the so-called greening agents, which

impart a vivid green color to a lawn, contain ammonium

sulfate. This salt is known to be particularly aggressive to

concrete. The relevance of the above statement is that

gardening procedures, including lawn and garden irrigation,

can greatly alter the sulfate conditions that existed at the

time of construction. Now, changes in the conditions in

service are beyond the scope of those involved in construc-

tion. Even if instruction handbooks and injunctions given to

homeowners do not allow the irrigation water to run off onto

foundations, such an approach is usually ineffective. Either

a complete moisture barrier around the foundations needs to

be provided (and this is not cheap or easy) or the home-

owner has to bear the consequences of his or her actions.

At the other extreme, once the soil has been compacted,

there may be limited availability of water, so that some

sulfates in the soil will not be available for reaction with

hydrated cement paste. According to Pye and Harrison [57],

‘‘a dry site where the depth of the water table is difficult to

find in any season is unlikely to give rise to significant

chemical attack on concrete placed on it’’.

3.4. My tests

I am not a newcomer to the study of sulfate attack on

concrete. In 1969, I published a paper about my tests on the

behavior of concrete in saturated and weak solutions of

magnesium sulfate [29]; the period of exposure was 1000

days. What is relevant to this paper is the type of tests that I

performed to assess the deterioration of concrete. These

were: a change in the weight of the specimen, a change in

the dynamic modulus of elasticity, and a change in the

length of the specimen. I did not use a determination of the

compressive strength of concrete because this is a destruc-

tive test so that, at each age, a different specimen would

have been tested, and having many dozen specimens for a

single mix would have been impractical. Compressive

strength can be determined on cores when the health of a

structure at a particular moment is at issue.

In my 1969 paper, I noted that ‘‘the changes in length,

weight, and frequency do not all show the same tendency, as

the influence of reactions of corrosion (sulfate attack) on the

three properties of concrete varies’’ [29]. This begs the

question: which test is the most appropriate?

The answer is that much depends on the mechanism

involved in the attack, which, in turn, depends on the cation

of the attacking sulfate. Calcium sulfate can result in the

formation of ettringite, which may (but not necessarily so)

lead to expansion. This can occur with sodium sulfate as

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961288

well, but sodium sulfate also leaches calcium hydroxide,

and this would result in a decrease in the dynamic modulus

of elasticity and, possibly, also in a decrease in weight.

I have couched the last paragraph in rather tentative

terms because, in a sulfate solution, just as much as in water,

cement continues to hydrate, and this initially improves the

properties of the specimen. It is only in the longer term that

the sulfate reactions dominate over the consequence of

continuing hydration.

Now, magnesium sulfate has a deleterious effect on

various properties of concrete and, specifically, on compres-

sive strength, in consequence of the decomposition of C-S-

H. However, if magnesium hydroxide (brucite) is allowed to

remain undisturbed on the surface of the concrete specimen,

a protective layer will be formed, and the sulfate attack will

be self-limiting.

In view of the above statements, I have to admit that my

tests were not structured as well as I would do now, but

then, hindsight has a 20/20 vision. The difference in the

action of various sulfates should be borne in mind; however,

on site, it is also possible for more than one type of sulfate to

be involved at the same time.

3.5. Change in compressive strength

If there is a dispute about whether concrete in a structure

has suffered damage due to sulfate attack, testing may be

necessary. It is well known that the presence of voids in

concrete decreases its compressive strength, the rule of

thumb being that 1% voids decrease the strength by about

5.5%. It is therefore logical to expect leaching of products of

hydration or the formation of cracks to cause a reduction in

the compressive strength. Thus, if cores taken from an

existing structure have a lower strength than the 28-day

strength at the time of placing the concrete, then it is

reasonable to conclude that the concrete has suffered dam-

age, possibly owing to sulfate attack.

This approach has been used on a number of occasions.

A few examples should suffice. Ibrahim et al. [58] wrote,

‘‘the effectiveness of these materials (surface treatment) in

decreasing the sulfate attack was evaluated by measuring the

reduction in compressive strength’’. Harrison [5] determined

the changes in compressive strength to establish whether

sulfate attack had taken place in 15-year long field exposure

tests. These findings were cited with approval by Figg [59].

In his laboratory studies, Brown [60] also used the loss of

compressive strength of mortar specimens to assess the

damage by sulfates. Al-Amoudi et al. [61] have also used

the change in compressive strength in their study of the

performance of blended cements exposed to magnesium and

sodium sulfates.

I would like to emphasize that what was measured in the

references reviewed above is the change in the compressive

strength consequent upon sulfate attack (or only exposure to

sulfates) and not the level of strength as such, which

depends on the mix used. I subscribe to the view that

determining the change in compressive strength is one

means of detecting sulfate attack; the absence of a reduction

in strength indicates that there has been no attack, even if

some reactions have taken place. It is unfortunate that I have

been wrongly listed as a reference in support of the view

that the determination of the compressive strength of cores

is irrelevant to proving that sulfate attack has or has not

taken place [62]. I am also misrepresented in the quotation

from my paper on high-alumina cement concrete by the

words: ‘‘strength, especially compressive strength, is an

inappropriate measure of durability’’ [6]. The common first

author of these two references is Skalny.

I am citing the above spurious reliance on my publica-

tions because the difference between the level of strength

and the change in strength is vital. What is significant with

reference to sulfate attack is that an increase in strength

(except at very early ages) is inconsistent with such an

attack, i.e., with deterioration of concrete by the action of

sulfates.

Laboratory tests on concrete exposed to sulfates are

generally performed on unstressed specimens, whereas, in

some service situations, stress may act concurrently with

sulfate attack. Laboratory tests on concrete under stress have

shown that ‘‘the durability of concrete under sodium sulfate

attack is affected by the state and level of sustained stress’’

[63]. However, the stress–strength ratio applied was gener-

ally higher than would be the case in practice. Nevertheless,

the effect of stress on damage during exposure to sulfates

represents an additional complication in the study of sulfate

attack.

3.6. ‘‘Layered damage’’ and tensile strength

In the absence of documents and clear-cut procedures for

the assessment of damage caused by sulfate attack, there is a

danger of one-off methods, sometimes quite fanciful, being

used to prove one’s point. Two of these may be worth

mentioning: the use of tensile strength (to the exclusion of

compressive strength) and the demonstration (real or al-

leged) of so-called layered damage.

It is to be expected that when an attack on concrete takes

place from one face of a concrete element, the extent of

damage will be progressive with the distance from the

exposed surface. This, for example, is observed in the case

of carbonation.

A much more complex situation has, however, been

proposed by some investigators who have introduced the

concept of layered damage, which implies that concrete

consists of a number of layers parallel to the exposed

surface, these layers having distinct properties. In conse-

quence, if a tensile force is applied at right angles to the

layers, failure will occur at a low tensile strength, while a

compressive force in the same direction would not be

affected by the layered nature of the damaged concrete.

A ‘‘proof’’ of the existence of layered damage was

provided by Ju et al. [64] by their measurement of ultrasonic

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1289

pulse velocity in different layers of cores taken from a

concrete slab. I have reviewed that work in a book [7] so

that only a brief mention here should suffice. The pulse

velocity measurements were taken at levels 15 mm apart on

four equally spaced diameters, and the average values of

velocity at each level were reported. On applying a statis-

tical analysis, I found that the mean value of all eight levels

lay within 95% confidence limits for one level, so that the

differences between the mean values at the different levels

were not statistically significant. Thus, the presence of

layers was not proved. Ju has not challenged my views.

Another ‘‘proof’’ of layer damage was proffered by an

expert by the use of the term ‘‘layers’’ by Taylor [32]. On a

closer study of his book [32], I found that by the words

‘‘surface layers may be progressively removed’’, he meant

laboratory preparation of successive surfaces for examina-

tion by X-ray diffractometry.

The assumption of layered damage has led to the

proposition that compressive strength tests are incapable

of detecting sulfate damage, but that the determination of

direct tensile strength does do so. I should make it clear

where I stand on this issue because my studies of concretes

suspected of having been subject to sulfate attack have

shown that the ratio of the splitting tensile strength to the

compressive strength of such concretes is the same as for

concrete stored under normal conditions [7]. I should point

out that I am referring to the splitting tensile strength, and

not to the direct tensile strength, because the latter is not

standardized by ASTM or by a British Standard, so that a

reliable test method is not available. For that matter, testing

concrete in direct tension is very difficult because of the

danger of eccentricity and of complex end stresses.

The suggestion that the compressive strength is unaffect-

ed by sulfate attack, while the splitting tension strength is

lowered, is founded upon a 1982 report by Harboe [65]. His

tests were performed on cores from the spillway of a dam in

Wyoming, built in the mid-1930s, the cores being tested in

1967. Harboe [65] found the compressive strength to be 40

MPa (5900 psi) while the tensile strength was only 2.2% of

that value. Normally, the tensile strength would be expected

to be about 10% of the compressive strength.

I have no explanation for those results, but they have not

been independently confirmed on any other concrete in the

succeeding 35 years. Harboe himself did not offer an expla-

nation for what appears to be an anomalous ratio of splitting

tensile strength to compressive strength. I find it difficult to

accept this anomaly because, if concrete has developed

cracks and contains damaged and soft hydrated cement paste,

how can its compressive strength remain unimpaired? Pro-

ponents of selective influence of alleged sulfate attack upon

mechanical properties of concrete have offered no answer to

my question about tensile and compressive strengths; my

view is that both compressive and tensile strengths must be

affected by damage in the same manner.

In addition to quoting Harboe [65], proponents of testing

concrete in direct tension also rely on Ju et al. [64], who say,

‘‘the uniaxial compression test is not indicative or suitable

for layer-damaged concrete cylinders’’ [64]; italics are in the

original text. Because the compression test is, according to

them, not suitable, they proceed to use direct tension testing,

basing themselves on ASTM D 2936-95, which is a test for

‘‘intact rock core specimens’’. Using standardized tests for

one material on a different material is a dubious practice.

Further work on tensile testing was done by Boyd and

Mindess [66], who interpreted their tests on concrete ex-

posed to sulfates to show that the splitting-tensile strength is

less sensitive to damage by sulfate. My interpretation of

their results is that they simply conform to the general

pattern of variation in the ratio of the tensile strength to

compressive strength: this ratio decreases with an increase

in strength. Thus, I do not see their tests as supporting the

hypothesis of the level of tensile strength being an indicator

of sulfate damage.

Moreover, Boyd and Mindess [66] introduced what they

describe as a ‘‘novel tensile strength test’’, actually devel-

oped in England in 1978, in which gas pressure is applied

to the curved surface of a cylinder. Unfortunately, as stated

by the inventor of the test, Clayton [67], the value of

strength so determined varies according to the choice of

loading medium. Perhaps this is why the new method died,

and resurrecting it 20 years later does not seem to be

profitable.

Locher [68] is cited as having used, as far back as 1966, the

tensile strength to determine sulfate resistance of mixes

containing slag. This is correct in that he tested 10� 10�60 mm mortar prisms exposed to various conditions [68].

However, given the shape and size of his specimens, the

determination of compressive strength was impossible. There

is no indication in his paper that he considered the measure-

ment of flexural strength preferable or superior to the com-

pressive strength: this was simply a necessity. Thus, I see

reliance on Locher [68] as a proponent of testing mortar (and

presumably concrete) in tension, as against testing in com-

pression, to be unfounded.

3.7. Use of scanning electron microscopy

This paper is not concerned with advanced testing

techniques, but a brief comment on the use of scanning

electron microscopy (SEM) may be in order. In some cases,

SEM has been used to establish that sulfate attack had

occurred by demonstrating the presence of ettringite. As I

understand the situation, SEM does not show the presence

of compounds but only of chemical elements, so that

compounds are inferred from the quantities of elements that

would be present in the given compound. Hence, the

presence of, say, ettringite cannot be unequivocally proven.

Moreover, the mere presence of ettringite is not a proof

of damage because much depends on whether the ettringite

is located in preexisting voids or cracks, or whether the

cracks were the consequence of expansion by ettringite

formation. Furthermore, Detwiler et al. [69] say that ‘‘ettrin-

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961290

gite is not a sign of sulfate attack unless its concentration is

greater than would be expected from hydration of cement’’.

It is not easy to establish how much ettringite originates

from the hydration of cement because, generally, the com-

position of the cement is not known at the time of the

investigation, which is usually several years after construc-

tion. Moreover, Detwiler et al. [69] point out that ‘‘in

concrete exposed to conditions of wetting and drying, it is

possible that sulfate concentration would be higher just

below the surface than immediately at the surface due to

washing out of ions’’. This is highly relevant to the fact that

SEM looks at an exceedingly small surface. Typically, using

a 1000-fold magnification, the area whose elemental anal-

ysis is determined would be 10� 6�10� 6 m and about

2�10� 6 m deep. For comparison, human hair has a typical

diameter of about 80�10� 6 m (but there is a decreasing

availability of hirsute men).

Because of the small size of the SEM picture, Detwiler et

al. [69] recommend the following procedure in the investi-

gation of suspected sulfate attack: ‘‘Step-by-step progres-

sion from large scale (tens of meters) to submicroscopic

scale (tens of nanometers) gave context to the testing,

allowing for selection of specimens that suitably represented

the concrete as a whole and for appropriate interpretation of

results’’. I find this to be a logical approach.

There is also an important influence of relative humidity

to which the test specimen is exposed in SEM (except in the

latest generation of instruments). This problem has been

very clearly considered by Kurtis et al. [70]. They point out

that ‘‘characterization techniques that require high vacuum

or drying. . .are not particularly appropriate as artifacts are

introduced’’ and go on to say that ‘‘methods, such as

scanning electron microscopy. . .have given limited and

misleading information about expansion reactions in con-

crete. . .’’ [70]. Accordingly, they recommend the use of soft

X-ray transmission microscopy, which does not require

drying or application of pressure [70]. Enthusiasts of SEM

might like to take note.

The danger of artifacts created by a high vacuum in the

sample chamber is pointed out also by Zelic et al. [71].

3.8. Mathematical modeling

A totally different approach to assessing the damage by

sulfates is by the use of mathematical models. They are

likely to become a valuable tool but, until they have been

calibrated and confirmed by field behavior—and we are not

there yet—they will not provide an understanding of the

physical phenomena involved. In addition, the boundary

conditions applied in a model may be arbitrary and conve-

niently simplistic.

This can happen with laboratory testing as well. Idorn

[72] criticizes laboratory procedures that confine the energy

conversion to occur at a constant temperature for reasons of

convenience, thus violating the true kinetics of the reactions

instead of modeling the reality of concrete.

It is pertinent to quote a statement made in the Magazine

of Concrete Research about site tests (for a different

purpose): ‘‘Conditions on site are not controlled in the same

way as in the laboratory, but the benefits of assessing real

effects at full scale are obvious, and allow greater confi-

dence in modeling and prediction procedures derived from

data presented and, by extrapolation, from data obtained by

others’’ [73]. It is also said about site data that ‘‘the benefits

of being able to confirm predictions made on the basis of

short-term or small-scale experiments are evident [73].

What I find highly surprising are mathematical models

that ‘‘predict’’ that failure has already occurred [64] and,

sometimes, even at a date preceding the construction of a

building. And yet, the building is still in use and has neither

been condemned nor recommended for urgent repairs. As an

engineer, I find a theory that is flatly contradicted by reality

too difficult to accept.

A problem with models is the choice of boundary

conditions, which sometimes may distort the output of a

model. For example, one theoretical analysis concluded that

‘‘numerical simulations indicate that the exposure to weak

sodium sulfate solutions may yield (sic) to a significant

reorganization of the internal microstructure of concrete’’

[74], This may be true, but it does not necessarily mean

sulfate attack and damage to concrete. Support for my

statement is given by DePuy [1] (referred to in the same

paper): ‘‘Chemical reactions per se in concrete are not

necessarily harmful’’.

Sometimes, the authors developing theoretical analyses

cite numerous references to support their assumptions, but

such support is diminished when the references are to tests

on concrete mixes of a bygone era, which used cements

significantly different from those used nowadays. To give

one example, the statement that ‘‘there is overwhelming

evidence to show that the degradation also contributes to a

significant reduction in the mechanical properties of con-

crete’’ [74] cites: a paper from the 1920s, another from the

1940s, two papers on laboratory tests, including neat cement

paste specimens, and a book by the proponent of the

theoretical analysis in question. I expect that the use of

‘‘self-support’’ is not all that uncommon; after all, every one

of us has confidence in our own publications. However,

there is the danger that a statement about ‘‘overwhelming

evidence’’ is likely to be cited again and may well become

embedded in the received wisdom, but is it really correct?

Another example of a sweeping generalization is a

statement that ‘‘none of them (building codes) contains

any specific limit concerning the maximum w/c that should

be selected for the production of concrete elements to be

exposed to negligible levels of sulfate. This is unfortunate

since, as emphasized by many authors, potentially destruc-

tive conditions may exist although analyses indicate the

ground water or soil to have a low sulfate content’’ [74]. The

references supporting the preceding statement include a

paper published in 1968, another published in 1982, and

one published in 1994.

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1291

I sometimes wonder whether some of the papers discus-

sing concrete in Southern California alleged to have been

attacked by sulfates have not been written to establish the

damage of those concretes. The models then developed are

premised on sulfate attack that had not been physically

established in an independent manner. I am not imputing ill

will because we can all be subject to what Bertrand Russell,

a great philosopher, called ‘‘the delusive support of subjec-

tive certainty’’.

There is a more frivolous, but nonetheless valid, support

of my view that an opinion repeated often enough, even if

incorrect, acquires currency. Lewis Carroll (a Cambridge

mathematician, best known as the author of Alice in Won-

derland ) wrote in The Hunting of the Snark, published in

1876: ‘‘What I tell you three times is true’’.

4. Part IV: An overview

4.1. Conclusions

Amid the plethora of confusing information and opinion,

some general conclusions can be drawn.

The term attack should be used only to describe damage

to concrete, and not just the fact that reactions have taken

place. Furthermore, the term sulfate attack should be ap-

plied only to the consequences of chemical reactions in-

volving the sulfate ion, recognizing that all reactions have

physical consequences. Damage by sulfates where the

action is purely physical and which can also be produced

by other salts should be termed physical attack.

Sulfate attack on concrete structures in service is not

widespread except in some areas, and the amount of

laboratory-based research seems to be disproportionately

large. On the other hand, our knowledge and understanding

of sulfate attack in the field remains inadequate.

The efficacy of Type V cement as against the use of a

low w/c in reducing sulfate attack on concrete is still

unresolved, but there is no doubt that dense concrete, well

compacted, and well cured is essential.

The significance of the cation in the sulfate is often not

appreciated, and yet, this influences the chemical reactions

and, thus, the consequences of sulfate attack.

The classification of the severity of exposure of concrete

to sulfates is not firmly established and has not been

calibrated by the determination of actual damage to concrete

under different conditions. A review of classifications used

in the past is given in Ref. [75].

The classification of exposure conditions should be clear

but also simple. It is not sensible to make elaborate and

complex recommendations for the selection of the concrete

mix when this classification is predicated on largely uncertain

and variable measurements of sulfate in the ground. We are

not dealing with an industrial product whose property can be

measured precisely, but rather with a variable natural mate-

rial—soil—that is furthermore disturbed by human opera-

tions during construction and by the user of the structure, who

may change the drainage or introduce salts while beautifying

or maintaining the land around the building.

Drainage of soil around the foundations is important

because dry salts do not attack the concrete.

The classification of severity by the determination of the

sulfate content in the soil is still a matter of debate. The

determination of the sulfate content in the soil should be

standardized in terms of the extraction ratio, so that the

outcome of the test cannot be varied according to the whim

or objectives of the tester. The amount of sulfate should be

expressed in milligrams of SO4 per kilogram of extract water.

The assessment of the extent of damage by the determi-

nation of the change in compressive strength appears to be

reliable. On the other hand, the argument that only tensile

strength testing should be used seems to be unfounded.

Mathematical models will be of help in understanding

sulfate attack, but so far, they have not been verified.

The use of SEM is not a primary evidence of sulfate

attack but should be preceded by tests on a large scale.

4.2. What next?

I would also like to draw some broader inferences. Alas, I

believe that the title of this paper is correct: Our knowledge

of sulfate attack is confused, and the practitioners in con-

struction lack a clear picture that would enable them to

achieve sulfate-resisting structures. Individual researchers

know much about specific actions and materials, but the

ensemble of their knowledge does not hold together. Unfor-

tunately, a reconciliation of these various opinions is not

possible at the present state of our knowledge. What this

paper has tried to achieve is to point out the conflicting

views, with reasons for the divergences in opinion where

these could be ascertained. Thus, we know at least why the

world of sulfate attack is confused.

The purpose of this paper was to present, as objectively

as I could, the problems in assessing the sulfate attack on

concrete and to point out the ramifications of generalized

statements about damage to concrete. To resolve the various

issues, much needs to be done. In other words, we need

research.

But what research? Limiting oneself to laboratory tests is

not enough. It is not particularly helpful to use accelerated

tests on miniscule specimens of neat cement paste or mortar.

Realistic conditions of exposure need to be achieved.

I would like to think that this paper will be of use in

showing where practical research is needed and is likely to

pay off. But this research has to be related to field conditions

of exposure to sulfates.

To say that researchers should be objective is to state the

obvious. But with large-scale litigation being the driving

force, all too often, research is undertaken to ‘‘prove a point’’

for one party or another, or to secure a ‘‘victory’’. Once a

researcher has become bound to his or her paymaster, even if

he or she wishes to remain objective, the researcher moves

Table 1

ACI 201.2R (1997) Guide to durable concrete. ACI 332R (1984, re-

approved 1997) Guide to Residential Concrete

Class of exposure Concentration of sulfates as SO4

In ground water (ppm) Water soluble in soil (%)

Mild < 150 < 0.10

Moderate 150–1500 0.10–0.20

Severe 1500–10,000 0.20–2.00

Very severe >10,000 >2.00

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961292

towards a goal and narrows the horizon of his or her work.

Moreover, even if this is not the case, he or she is perceived to

be biased, and this prompts others working on the topic to

prove the researcher wrong. And so it goes on.

I am not saying the above to criticize anybody because

we are all fallible ‘‘sinners’’. What we have to beware of,

however, is that sweeping assertions that are advanced,

without objective and verifiable documentation available

for all to see, are not good science. These assertions should

not become a part of the body of knowledge upon which all

concrete specialists in the world at large can rely.

I am not recommending a mammoth experimental inves-

tigation. I believe that much can be gleaned from revisiting

published information, both on laboratory test results and

on-site observations, as well as from tests on field concrete.

This needs to be done by a team of chemists, petrographers,

and structural engineers, all working together.

Thomas Macaulay, an early 19th century historian, wrote

in How Horatius Kept the Bridge:

‘‘Now who will stand on either hand and keep the bridge

with me’’

The leader of the team should be a new person versed in

matters structural, and he or she should be flanked by

someone (on the right hand) versed in chemistry of concrete,

not just cement, and by someone (on the left hand) expe-

rienced in optical microscopy of field concrete.

If eyebrows are raised at the inclusion of structural

engineers, my answer is that the objective of establishing

facts about sulfate attack and, hence, providing guidance on

achieving sulfate-resisting structures, is to build structures

that will be durable for the required service life under the

expected conditions of exposure.

I have used the term ‘‘sulfate-resisting structures’’ rather

than sulfate-resisting concrete because, all too often, the

field of vision of researchers is limited to the material,

whereas the purpose of studying concrete is to build good

and durable structures. If we do not succeed in this,

competitors to concrete structures—and we must not delude

ourselves by thinking that concrete is irreplaceable—will

gain the ascendancy. But then, I am a structural engineer and

not a backroom researcher; I offer no apology for this.

At the outset of this paper, I said that sulfate has generally

not caused significant damage to structures in service. Sowhy

this major discussion and recommendations for an indepen-

dent review and study? The answer is that a lack of clarity

permits some people to claim sulfate attack, and, at present,

there is no reliable and coherent body of knowledge that

would make it easy to distinguish real damage from spurious

allegations. The reaction of some designers has already been

to overprovide protective measures so as to avoid potential

(even if unjustified) claims in the future. This costs the client

money and reduces the competitiveness of concrete.

It would be unsatisfactory for this paper to end by

observing simply that our knowledge of sulfate attack on

structures in service is confusing. On the other hand, we

are not yet in a position to present a coherent and

complete picture on the basis of which we could select

suitable concrete mixes and design durable structures, as

well as assess objectively the state and probable future of

existing structures. To achieve this, we need to review all

the available information and to pull it together, clearly

distinguishing laboratory results and field observations. It

is the latter—structural behavior—that is of paramount

importance, but scientific knowledge is also necessary.

Finally, I would like to make a very general remark:

There is an incompatibility between the breadth of spec-

ified limits on the properties of cement on the one hand,

and on the other, the more demanding expectations of

concrete, which can be achieved by the use of cementi-

tious materials with specific properties. Now, portland

cement needs to comply only with very broad limits on

strength, fineness, and chemical compounds. At the same

time, requirements such as compatibility with superplasti-

cizers or ensuring sulfate resistance need close and

narrow controls. Whenever this issue is raised, cement

manufacturers reply that you cannot have a high-tech

material for the price of a low-tech one. Which do we

want? Consideration of this dilemma requires a separate

study.

An essential part of sulfate is sulfur, and the element

sulfur has long been known to contribute to damage. In

King Lear, Shakespeare says:

‘‘There’s hell, there’s darkness,

there is the sulphurous pit,

Burning, scalding, stench, consumption;

fie, fie, fie!’’

and also

‘‘Singe my white head!’’

which, given my white head, is an appropriate ending to this

paper.

Appendix A. Classification of exposure to sulfates

This Appendix contains a selection of tables from

American, Canadian, British, and European sources giv-

ing a classification of exposure to sulfates (Tables 1–15).

Table 2

ACI 318 (1983 through 2002) Building Code

Class of exposure Concentration of sulfates as SO4

In water (ppm) Water soluble in soil

(percent by weight)

Negligible < 150 < 0.10

Moderate 150–1500 0.10–0.20

Severe 1500–10,000 0.20–2.00

Very severe >10,000 >2.00

Table 3

U.S Bureau of Reclamation Concrete Manual (1966)

Class of exposure Concentration of sulfates as SO4

In water (ppm) Water soluble in soil (%)

Negligible 0–150 < 1000

Positive 150–1000 1000–2000

Considerable 1000–2000 2000–5000

Severe >2000 >5000

Table 4

Uniform Building Code (1991)

Class of exposure Concentration of sulfates as SO4

In ground water (ppm) Water soluble in soil

(percent by weight)

Negligible < 150 < 0.10

Moderate 150–1500 0.10–0.20

Severe 1500–10,000 0.20–2.00

Very severe >10,000 >2.00

Table 5

Canadian Standard A23.1.94

Class of exposure Concentration of sulfates as SO4

In ground water (ppm) Water soluble in soil (%)

S-3 moderate 150–1500 0.10–0.20

S-2 severe 1500–10,000 0.20–2.0

S-1 very severe >10,000 >2.0

Table 6

European Standard EN206-1:2000

Class of exposure Concentration of sulfates as SO4 Concentration of

(chemical

environment)In ground

water (ppm)

In soil

(ppm total)

magnesium in

ground water (ppm)

Slightly

aggressive

200–600 2000–3000 300–1000

Moderately

aggressive

600–3000 3000–12,000 1000–3000

Highly

aggressive

3000–6000 12,000–24,000 3000 to saturation

The extraction of SO4 from soil is by hydrochloric acid. However, water

extraction may be used if experience is available in the place of use of the

concrete.

Clay soils with a permeability below 10� 5 m/s may be moved into a lower

class.

Table 7

British Building Research Station: notes on concrete in sulfate-bearing

clays and ground waters, The Builder, July 7, 1939, p.29. This table was

incorporated into British BRE Digest 31 (1951)

Class of exposure Concentration of sulfates as SO3

In water (ppm) In clay

(percent by weight)

1 < 300 < 0.2

2 300–1000 0.2–0.5

3 >1000 >0.5

Analysis of ground water is preferred to that of clay.

In some cases, it may be found that the results of the ground water and clay

analyses lead to different classifications; the more severe classification

should be adopted.

Table 8

British Code of Practice CP 110:1972

Class of Concentration of sulfates as SO3

exposureIn ground

water (ppm)

In 2:1

water– soil

extract (g/l)

Total in soil

(ppm)

1 < 300 – < 2000

2 300–1200 – 2000–5000

3 1200–2500 1.9–3.1 5000–10,000

4 2500–5000 3.1–5.6 10,000–20,000

5 >5000 >5.6 >20,000

For classes 3 to 5, if calcium sulfate is predominant, the water–soil extract

may give a lower class of exposure than the total SO3.

To convert SO3 to SO4, multiply by 1.2.

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1293

These tables show that it is not only the sulfate content

boundaries of the various exposure classes that vary, but

also the water–soil extraction ratios (sometimes unspeci-

fied), as well as the method of determining the sulfate

content: acid or water soluble. Furthermore, the concen-

tration of the sulfate in the soil is sometimes expressed per

unit mass of soil; in other cases, per unit mass of

extraction water. The rationale of these various approaches

is not self-evident.

Even a quick look at the ensemble of the classifications

of exposure conditions and at the class boundaries shows

substantial differences. However, their full significance can

Table 9

British Code of Practice CP 8110:1985

Class of Concentration of sulfates as SO3

exposureIn ground

water (g/l)

In 2:1

water–soil

extract (g/l)

Total (ppm)

1 < 0.3 < 1.0 < 2000

2 0.3–1.2 1.0–1.9 2000–5000

3 1.2–2.5 1.9–3.1 5000–10,000

4 2.5–5.0 3.1–5.6 10,000–20,000

5 >5.0 >5.6 >20,000

If calcium sulfate is predominant the water– soil extract may give a lower

class than total SO3.

To convert SO3 to SO4, multiply by 1.2.

Table 13

British BRE Digest 250 (1981)

Class of Concentration of sulfates as SO3

exposureIn ground

water (ppm)

In 2:1 water– soil

extract (g/l)

Total (ppm)

1 < 300 < 1.0 < 2000

2 300–1200 1.0–1.9 2000–5000

3 1200–2500 1.9–3.1 5000–10,000

4 2500–5000 3.1–5.6 10,000–20,000

5 >5000 >5.6 >20,000

To convert SO3 to SO4 , multiply by 1.2.

Table 14

British BRE Digest 363 (1991)

Class of

exposure

Concentration of sulfate as SO4 Concentration of

magnesium

In ground In soil In ground In soil

water (ppm)By acid

extraction

(%)

By 2:1

water– soil

extract

water

(ppm)

(g/l)

Table 10

British Standard BS 5328:1997 Guide to specifying concrete

Class of

exposure

Concentration of sulfate as SO4 Concentration of

magnesium

In ground In soil In ground In soil (g/l)

water

(ppm)By acid

extraction

(%)

By 2:1

water– soil

extract (g/l)

water

(ppm)

1 < 400 < 0.24 < 1.2 – –

2 400–1400 * 1.2–2.3 – –

3 1500–3000 * 2.4–3.7 – –

4A 3100–6000 * 3.8–6.7 V 1000 V 1.2

4B 3100–6000 * 3.8–6.7 >1000 >1.2

5A >6000 * >6.7 V 1000 V 1.2

5B >6000 * >6.7 >1000 >1.2

Consider both SO4 and magnesium.

Classification on the basis of groundwater is preferred.

Higher values are given for the water–soil extract in recognition of the

difficulty of obtaining representative samples and of achieving a

comparable extraction rate to that from ground water.

* Classify on the basis of 2:1 water– soil extract.

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961294

be assessed only by a consideration of the properties of the

concrete mixes corresponding to the various exposure

classes.

Such a consideration is not possible because the proper-

ties of the various cements were not the same in the different

countries. Moreover, the actual properties of cements varied

over time, and the standard requirements for cements were

also not immutable.

I am not saying that such a study is impossible, but it

would be a very major undertaking. Moreover, the study

Table 11

British BRE Digest 90 (1968)

Class of Concentration of sulfates as SO3

exposureIn ground

water (ppm)

In 2:1 water–soil

extract (g/l)

Total (ppm)

1 < 300 – < 2000

2 300–1200 – 2000–5000

3 1200–2500 2.5–5.0 5000–10,000

4 2500–5000 5.0–10.0 10,000–20,000

5 >5000 >10.0 >20,000

To convert SO3 to SO4, multiply by 1.2.

Table 12

British BRE Digest 174 (1975)

Class of Concentration of sulfates as SO3

exposureIn ground

water (ppm)

In 2:1

water–soil

extract (g/l)

Total (ppm)

1 < 300 – < 2000

2 300–1200 – 2000–5000

3 1200–2500 1.9–3.1 5000–10,000

4 2500–5000 3.1–5.6 10,000–20,000

5 >5000 >5.6 >20,000

To convert SO3 to SO4 , multiply by 1.2.

would be meaningful only if we had a record of long-term

performance of the various concretes under the actual

exposure conditions. Alas, with very few exceptions of

long-term field studies, one in California and one in Eng-

land, no such information is available.

The preceding discussion exemplifies our dilemma: we

do not know whether the exposure class boundaries are

‘‘correct’’ because they have not been calibrated or verified.

This then is the reason for the paper titled, ‘‘The confused

(g/l)

1 < 400 < 0.24 < 1.2

2 400–1400 * 1.2–2.3

3 1400–3000 * 2.3–3.7

4a 3000–6000 * 3.7–6.7 < 1.0 < 1.2

4b 3000–6000 * 3.7–6.7 >1.0 >1.2

5a >6000 * >6.7 < 1.0 < 1.2

5b >6000 * >6.7 >1.0 >1.2

* If >0.24, classify on the basis of 2:1 water–soil extract.

Table 15

British BRE Special Digest 1 (2002). British Standard BS 8500:2000

Class of Concentration of sulfate as SO4 Concentration of magnesium

exposureIn ground

water (ppm)

In 2:1

water– soil

extract (g/l)

In ground

water (ppm)

In 2:1

water– soil

extract (ppm)

DS-1 < 400 < 1200

DS-2 400–1400 1200–2300

DS-3 1500–3000 2400–3700

DS-4a 3100–6000 3800–6700 V 1000 V 1200

DS-4b 3100–6000 3800–6700 >1000 >1200

DS-5a >6000 >6700 V 1000 V 1200

DS-5b >6000 >6700 >1000 >1200

Consider both SO4 and magnesium.

Static and mobile water are distinguished.

The value of pH of the water is also considered.

A. Neville / Cement and Concrete Research 34 (2004) 1275–1296 1295

world of sulfate attack’’ and this is why the entire arena of

sulfate attack in the field needs to be studied.

References

[1] G.W. DePuy, Chemical resistance of concrete, Concrete and Con-

crete—Making Materials, ASTM STP 169C, Philadelphia, PA,

1994, 263–281.

[2] W.G. Hime, Chemists should be studying chemical attack on concrete,

Concr. Int. 25 (4) (2003) 82–84.

[3] N. Crammond, The occurrence of thaumasite in modern construc-

tion—a review, Cem. Concr. Compos. 24 (3–4) (2002) 393–402.

[4] J. Bensted, Thaumasite—three-year review, Concrete, (2003 Sept.)

7.

[5] W.H. Harrison, Sulphate resistance of buried concrete, Building Re-

search Establishment Report, 1992.

[6] J.P. Skalny, I. Odler, J. Marchand, Sulfate Attack on Concrete, Spon,

London, 2001.

[7] A. Neville, Neville on Concrete, ACI, Farmington Hills, MI, USA,

2003.

[8] P.K. Mehta, Sulfate attack on concrete: separating myths from reality,

Concr. Int. 22 (8) (2000) 57–61.

[9] Concrete Society, Diagnosis of Deterioration in Concrete Structures,

Technical Report, No.54, 2000.

[10] M. Collepardi, A state-of-the art review on delayed ettringite attack on

concrete, Cem. Concr. Compos. 25 (2003) 401–407.

[11] J. Skalny, I. Odler, F. Young, Discussion of Ref. 12, Cem. Concr. Res.

30 (2000) 161–162.

[12] W.G. Hime, B. Mather, ‘Sulfate attack’, or is it? Cem. Concr. Res. 29

(1999) 789–791.

[13] H. Lafuma, Theorie de l’expansion des liants hydrauliques, La revue

de materiaux de construction et de travaux publics, Part 1, 243 (1929

Dec.) 441–444 and Part 2, 244 (1930 Jan.) 4–8.

[14] D.G.R. Bonnell, M.E. Nottage, Studies in porous materials with spe-

cial reference to building materials: I. The crystallisation of salts in

porous materials, J. Soc. Chem. Ind. 58 (1939) 16–21.

[15] F.M. Lea, R.W. Nurse, Problems of crystal growth in building materi-

als, Discuss. Faraday Soc. 5 (1949) 345–351.

[16] P.W. Brown, H.F.W. Taylor, The role of ettringite in external sulfate

attack, Sulfate Attack Mechanisms, Materials Science of Concrete,

Amer. Ceramic Society, Ohio, 1999, 73–97.

[17] P.K. Mehta, Discussion, Sulfate Attack Mechanisms, Materials Sci-

ence of Concrete, Amer. Ceramic Society, Ohio, 1999, 22–23.

[18] R.J. Flatt, Salt damage in porous materials: how high supersaturations

are generated, J. Cryst. Growth 242 (2002) 435–454.

[19] P.W. Brown, Thaumasite formation and other forms of sulfate attack,

guest editorial, Cem. Concr. Compos. 24 (3–4) (2002) 301–303.

[20] A. Neville, Concrete: 40 years of progress?Concrete 38 (2004 Feb.)

52–54.

[21] P.K. Mehta, Sulfate attack on concrete: a critical review, Materials

Science of Concrete, vol. III, Amer. Ceramic Society, Ohio, 1993,

105–130.

[22] J.S. Pierce, Discussion, Sulfate Attack Mechanisms, Materials Science

of Concrete, Amer. Ceramic Society, Ohio, 1999, 41–43.

[23] R.E. Tobin, Sulfate Reactions in Concrete, Important Technical

Information, (2000 May 22).

[24] H.F.W. Taylor, Discussion, Sulfate Attack Mechanisms, Materials Sci-

ence of Concrete Special Volume, Amer. Ceramic Society, Ohio,

1999, 33–34.

[25] M. Santhanam, M.D. Cohen, J. Olek, Sulfate attack research—whither

now?, Cem. Concr. Res. 31 (2001) 845–851.

[26] B. Tian, M.D. Cohen, Expansion of alite paste caused by gypsum

formation during sulfate attack, J. Mater. Civ. Eng. 12 (1) (2000)

24–25.

[27] B. Mather, Discussion of The process of sulfate attack on cement

mortars by Shen Yang, Xu Zhongzi, and Tang Mingshu, Adv. Cem.

Based Mater., (1997) 109-111.

[28] Rasheeduzzafar, O.S.B. Al-Amoudi, S.N. Abduljauwad, M. Masle-

huddin, Magnesium– sodium sulfate attack in plain and blended

cements, ASCE J. Mater. Civ. Eng. 6 (2) (1994) 201–222.

[29] A.M. Neville, Behavior of concrete in saturated and weak solutions of

magnesium sulphate or calcium chloride, ASTM J. Mater. 4 (4) (1969)

781–816.

[30] V.N. Shah, C.J. Hookham, Long-term aging of light water reactor

concrete containments, Nucl. Eng. Des. 185 (1998) 51–81.

[31] M. Eglinton, Resistance of concrete to destructive agencies, Lea’s

Chemistry of Cement, Fourth edition, Arnold, London, 1998,

299–342.

[32] H.F.W. Taylor, Cement Chemistry, Second edition, Thomas Telford,

London, 1997.

[33] Bureau of Reclamation Concrete Manual, Eighth edition, U.S. Gov-

ernment Printing Office, Washington, DC, 1985.

[34] Concrete Society Working Party, The influence of cement content on

the performance of concrete, Concr. Soc. Discuss. Doc., CS 125, 1999.

[35] F.M. Lea, The Chemistry of Cement and Concrete, Third edition,

Arnold, London, 1970.

[36] B. Mather, Field and laboratory studies of the sulphate resistance of

concrete in Performance of Concrete, Univ. of Toronto Press, Toronto,

1968.

[37] T. Thorvaldson, Chemical aspects of the durability of cement prod-

ucts, Third International Symposium on Chemistry of Cement, Ce-

ment and Concrete Assn., London, England, 1954, 436–484.

[38] O.S.B. Al-Amoudi, Attack on plain and blended cements exposed to

aggressive sulfate environments, Cem. Concr. Compos. 24 (3–4)

(2002) 305–316.

[39] E.F. Irassar, M. Gonzalez, V. Rahhal, Sulphate resistance of Type V

cements with limestone filler and natural pozzolana, Cem. Concr.

Compos. 22 (2000) 361–368.

[40] D. Stark, Durability of concrete in sulfate-rich soils, Research and

Development Bulletin, vol. RD O97, Portland Cement Assn., Skokie,

IL, 1989.

[41] D. Stark, Performance of concrete in sulfate environments, Research

and Development Bulletin, vol. RD 129, Portland Cement Assn.,

Skokie, IL, 2002.

[42] BRE Special Digest 1, Concrete in Aggressive Ground, Parts 1–4,

second edition, BRE Centre for Concrete Construction and Centre for

Ground Engineering Remediation, 2003.

[43] British Standard BS 5328: Pt. 1, Specifying Concrete, 1997.

[44] R.P. Khatri, V. Sirivivatnanon, J.L. Yang, Role of permeability in

sulphate attack, Cem. Concr. Res. 27 (8) (1997) 1179–1189.

[45] S.P. Shah, K. Wang, W.J. Weiss, Mix proportioning for durable con-

crete, Concr. Int. 12 (9) (2000) 73–78.

[46] J.F. Young, S. Mindess, R.J. Gray, A. Bentur, The Science and Tech-

nology of Civil Engineering Materials, Prentice-Hall, New Jersey,

1998.

[47] K.E. Kurtis, P.J.M. Monteiro, S.M. Madanat, Empirical models to

predict concrete expansion caused by sulfate attack, ACI Mater. J.

97 (2) (2000) 156–167 (and Erratum 97 (6) (2000) 713).

[48] P.J.M. Monteiro, K.E. Kurtis, Time to failure for concrete exposed to

severe sulfate attack, Cem. Concr. Res. 33 (7) (2003) 987–993.

[49] D.J. Corr, P.J.M. Monteiro, K.E. Kurtis, A. Der Kiureghian, Sulfate

attack of concrete: reliability analysis, ACI Mater. J. 98 (2) (2001)

99–104.

[50] S.J. Ficcadenti, Effects of cement type and water to cement ratio on

concrete expansion caused by sulfate attack, Struct. Eng., Mech. Com-

put. 2 (2001) 1607–1613.

[51] British Standard BS 8110, Structural Use of Concrete, 1985.

[52] European Standard EN 206-1, Concrete-Specification, Performance,

Production and Conformity, 2000.

[53] ACI Building Code, 318, 2002.

[54] P. Nixon, Concrete in aggressive ground: an introduction to BRE

Special Digest 1, Concrete, (2001 Sept.) 62–64.

A. Neville / Cement and Concrete Research 34 (2004) 1275–12961296

[55] California Department of Transportation, Test 417, Method of Testing

Soils and Waters for Sulfate Content, 1986.

[56] J. Marchand, E. Samson, Y. Maltais, R.J. Lee, S. Sahu, Predicting the

performance of concrete structures exposed to chemically aggressive

environment-field validation, Mat. Struct. 35 (2002 Dec.) 623–631.

[57] P.W. Pye, H.W. Harrison, Performance, diagnosis, maintenance, repair

and avoidance of defects, BRE Building Elements: Floors and Floor-

ing, 1997.

[58] M. Ibrahim, A.S. Al-Gahtani, M. Maslehuddin, F.H. Dakhil, Use of

surface treatment materials to improve concrete durability, ASCE J.

Mater. Civ. Eng. 11 (1) (1999) 36–40.

[59] J. Figg, Field studies of sulfate attack on concrete, Sulfate Attack

Mechanisms, Materials Science of Concrete, Amer. Ceramic Society,

Ohio, 1999, 315–323.

[60] P.W. Brown, An evaluation of the sulfate resistance of cements, Cem.

Concr. Res. 11 (1981) 719–727.

[61] O.S.B. Al-Amoudi, M. Maslehuddin, M.M. Saadi, Effect of magne-

sium sulfate and sodium sulfate on the durability performance of plain

and blended cements, ACI Mater. J. 92 (1) (1995) 15–24.

[62] J. Skalny, J.S. Pierce, Sulfate attack: an overview, Sulfate Attack

Mechanisms, Materials Science of Concrete, Amer. Ceramic Society,

Ohio, 1999, 49–63.

[63] U. Schneider, W.G. Piasta, The behaviour of concrete under Na2SO4

solution attack and sustained compression or bending, Mag. Concr.

Res. 43 (157) (1991) 281–289.

[64] J.W. Ju, L.S. Weng, S. Mindess, A.J. Boyd, Damage assessment of

service life prediction of concrete subject to sulfate attack, Sulfate

Attack Mechanisms, Materials Science of Concrete, Amer. Ceramic

Society, Ohio, 1999, 265–282.

[65] E.M. Harboe, Longtime studies and field experiences with sulfate

attack, George Verbeck Symposium on Sulfate Resistance of Con-

crete, SP-77, ACI, Farmington Hills, MI, 1982, 1–18.

[66] A.J. Boyd, S. Mindess, The effect of sulfate attack on the tensile to

compressive strength ratio of concrete, Third International Conference

on Concrete Under Severe Conditions, ACI/CSCE, Vancouver, 2001,

789–796.

[67] N. Clayton, Fluid-pressure testing of concrete cylinders, Mag. Concr.

Res. 30 (102) (1978) 26–30.

[68] F.W. Locher, Zur Frage des Sulfatwiderstands von Huttenzementen

(The problem of the sulfate resistance of slag cements), Zem.-Kalk-

Gips 9 (1966) 395–401.

[69] R.J. Detwiler, P.C. Taylor, L.J. Powers, W.G. Corley, J.B. Delles, B.R.

Johnson, Assessment of concrete in sulfate soils, ASCE J. Perform.

Constr. Facil. 14 (3) (2000) 89–96.

[70] K.E. Kurtis, P.J.M. Monteiro, J.I. Brown, W. Meyer-Ilse, Expansive

reactions in concrete observed by soft X-ray transmission microscopy,

Mater. Res. Soc. Symp. Proc. 524 (1998) 3–9.

[71] J. Zelic, R. Krstulovic, E. Tkalcec, P. Krolo, Reply to the discussion of

the paper ‘Durability of the hydrated limestone-silica fume Portland

cement mortars under sulphate attack’, Cem. Concr. Res. 30 (2000)

333.

[72] G.M. Idorn, Sulphate attack on concrete, Sulfate Attack Mechanisms,

Materials Science of Concrete, Amer. Ceramic Society, Ohio, 1999,

65–72.

[73] J.N. Clarke, Editorial, Mag. Concr. Res. 40 (145) (1988) 194.

[74] E. Samson, Y. Maltais, J.J. Beaudoin, J. Marchand, Theoretical anal-

ysis of the effect of weak sodium sulfate solutions on the durability of

concrete, Cem. Concr. Compos. 24 (3–4) (2002) 317–329.

[75] I. Longworth, Development of guidance on classification of sulfate-

bearing ground for concrete, Concrete 38 (2004 Feb.) 25–26.